To translate scanned documents and images, you need a tool that can recognize the printed text before translating it. This guide walks through Lara Translate’s PDF and image workflows, using a delivery note and an instruction sheet to show which settings to choose and what to check in the downloaded files.

How Does Scanned Document Translation Work?

Scanned document translation uses optical character recognition (OCR) to identify printed text in an image, translates that text into another language, and reconstructs the output file. You can translate scanned documents without typing out each page, but you still need to check that recognition, translation, and layout processing worked correctly.

Check which type of file you have before choosing a workflow.

Try selecting a sentence in your PDF. If you cannot, the page may be image-based. If you can, still check the rest of the file: a selectable paragraph does not mean a screenshot on the next page has selectable text.

For more information, you can check our guide to what OCR is and how it helps with translation.

How to Prepare Your Scanned Document Before Translation

Before you translate scanned documents, make sure every page is complete and readable. Check that the tool supports your file and language pair, keep an untouched original, and confirm you can upload the content under your organization’s rules.

Improve Image Clarity and Page Orientation

Open the file and zoom in on the smallest text you need translated. If you struggle to read it, start by improving the source. For a photograph, place the page flat, keep the camera parallel to it, and avoid shadows or reflections across the words. Include the full page so headings, footnotes, and final lines remain visible.

Rotate sideways pages and straighten tilted scans. Keep the clearest original available. Repeated compression can make thin letters and decimal points harder to recognize. Enlarging a blurry image will not restore missing detail. If you have the paper document, a new scan or photograph may be more useful.

Check File Support, Languages, and Upload Requirements

Check the exact file extension and target language before starting. Lara Translate supports JPG, JPEG, PNG, TIFF, and other image formats. You can check the image-to-image support guide for the languages available in this workflow before uploading your file.

If the original editable document is available, consider translating that instead of its scan. A source Word file gives the tool text and document structure directly. Use the scanned-file workflow when the scan or image is the version you have.

Check the file size, page, and usage limits shown in your account before uploading a batch. Document and image translation use your character allowance differently.

For a standalone image, Lara Translate consumes a fixed amount of 1,500 characters per translated image, regardless of how much text the image contains. For images inside a PDF, usage is based on the number of characters extracted from and translated in each image.

Review Lara Translate’s usage guidance before processing several files.

Keep the Original and Check Confidentiality Requirements

Save the source separately from the translated output. Use filenames that identify the document and language, such as delivery-note-original.pdf and delivery-note-en.pdf. This makes comparison easier and prevents someone from mistaking the translation for the source.

Before uploading business records, confirm that the service and account settings meet your organization’s requirements. An invoice or delivery note may contain customer addresses, account details, or commercially sensitive information.

How to Translate a Scanned Document with Lara Translate

To translate a scanned document with Lara Translate, upload it in the Documents tab, confirm the languages, and start the translation. Download the document file translation result and compare it with the original, checking that product descriptions, quantities, and instructions remain complete and correctly positioned.

Step 1: Upload Your Scanned Document

Open Lara Translate and go to Documents. Drag your PDF into the upload area or use Browse files to select it.

Wait for the upload and analysis to finish. Check that you selected the right document and review the detected source language. If the file has several pages, open the original separately and confirm that it contains the complete delivery note.

Step 2: Enable Translate Images and Choose the Target Language

Confirm the source and target languages before submitting the file, particularly if the scan contains little text or several languages.

If you are translating several files, check the filenames before starting the run. A revised delivery note may look almost identical to the previous version while containing a different quantity or delivery date.

The same setting also matters for a PDF containing ordinary text and embedded screenshots. Our guide to translating images inside PDFs covers those mixed documents in more detail.

Step 3: Translate and Download the File

Start the translation and wait for processing to complete. Use the Download icon to save the translated result, then open the file on your computer.

Check the output format and page count. For the delivery note, compare the product rows, quantities, and instructions before sending it to the receiving team. Do not treat a completed download as confirmation that every detail is correct.

How to Translate an Image Document with Lara Translate

To translate a JPG or PNG document with Lara Translate, upload the image, choose the target language, and start image translation. Download the translated image in its original format and compare the wording and text placement with the source. You do not need to convert the image into a PDF first.

Step 1: Upload Your Scanned Image or Photograph

Visit the Lara Translate image translation page. Upload the JPG or PNG you want to translate.

Check that the image contains the complete instruction sheet. Look especially at text near the edges, small labels, and any warnings printed below the main instructions. If these are cut off or unreadable, correct the source before translating.

Step 2: Choose the Target Language and Start Translation

Choose the target language and confirm the source language. Start the translation.

Lara Translate recognizes the text, translates it, and reconstructs the image with the translated wording. The placement and layout are preserved as much as possible, but the result still needs checking, especially around small text and crowded sections.

Step 3: Download the Translated Image

Save the translated output and open it beside the original. Lara Translate provides same-format output for standalone images, such as JPG to JPG and PNG to PNG.

A translated image is not automatically an editable Word document. If you need to revise the wording extensively or reuse it in another design, plan that editing step separately.

Read the translated instructions in order. Check that each heading still belongs to the correct paragraph and that warnings remain easy to find.

For example, if the sheet says to store a product between 5°C and 25°C, check both numbers and the unit. Then read any instruction beginning with “do not” against the source. A missing negative can change what someone does, even when the surrounding text sounds natural.

How to Check the Translated Document Before Sharing

You can check a translated scan by comparing it with the original page by page. Verify names, codes, numbers, dates, and instructions first, then inspect tables, reading order, and text placement. Ask a qualified reviewer to check content where a mistake could affect a decision, obligation, or safety instruction.

Verify Names, Numbers, Dates, and Terminology

Compare details that must remain accurate: company names, product codes, quantities, currencies, measurements, and dates. OCR can confuse similar characters, so check codes against the source rather than relying on whether they look plausible.

For the delivery note, read across each row. Confirm that the product description still matches its quantity and unit. Check decimal separators and date formats for the target audience, too. A translated document should not introduce uncertainty about how much was delivered or when.

Review recurring terms consistently. If the same component has two different names in the translation, resolve the wording before circulating the file.

Compare Tables, Reading Order, and Text Placement

Look at the whole page, then inspect individual sections. Check that columns remain separate, table entries align with the right headings, and translated paragraphs do not cover nearby content.

Small print needs its own check. Footnotes, labels, and instructions at the bottom of a page can be missed when you focus only on the main paragraphs. Longer translated wording may also wrap differently or become harder to read.

Arrange Additional Review Where Needed

Choose the review level according to what people will do with the translation. A document used to understand a routine message needs a different approval process from a safety instruction, signed agreement, or official submission.

Ask a reviewer who understands the language and subject to check consequential content. If the recipient requires a certified translation, confirm those requirements before submitting an AI-translated file.

For sensitive or business-critical documents, Lara Translate’s AI + Human translation service offers review by a professional linguist from Translated. For certified, sworn, or official submissions, confirm that the requested service meets the recipient’s requirements.

Common Scanned Document Translation Problems and How to Fix Them

If a scanned document translates incorrectly, check the source image, the image translation setting, and the affected layout before trying again.

Conclusion

The practical way to translate scanned documents or images is to start with one representative file, use the workflow that matches its format, and review the downloaded result before processing the rest. Enable image translation for a scanned PDF and use the image workflow for a standalone JPG or PNG.

Lara Translate brings text recognition and translation into the same process, so you can translate scanned documents without extracting and pasting the words into another tool. Start with a clear scan, select the appropriate settings, and check the downloaded file against the original before sharing it.

Frequently Asked Questions

How do you translate scanned PDFs?

To translate scanned PDFs, use a tool with OCR to recognize the printed text before translating it. In Lara Translate, open Documents, upload your PDF and select the target language. Click Translate, download the result, and check the wording, numbers, and layout against the original.

How much quota does Lara Translate use for image translation?

Lara Translate uses a fixed allowance of 1,500 characters for each standalone image, regardless of how much text the image contains. For images inside a PDF, usage depends on the number of characters extracted from and translated in each image. The PDF’s total usage includes both its regular text and translated image text. Check your remaining character allowance before translating multiple images or PDFs.

How do you translate a PDF file from French to English?

To translate a PDF file from French to English, upload the PDF to Lara Translate’s Documents tab, confirm French as the source language, and select English as the target. Start the translation, download the file, and compare important details with the French original.

Can Google Translate translate a scanned document?

Google Translate’s Documents feature does not translate text inside images or scanned PDF pages. According to Google’s Help documentation, this text may appear in the output document but remains untranslated. If you have a scanned PDF, use an OCR tool to extract its text before translating it, or use an image translation tool for individual page images.

How do you translate a PDF on Mac?

To translate selected text in a PDF on Mac, open the file in Preview, select a sentence or paragraph, then right-click and choose Translate [text]. Select your target language to read the translation, or click Copy Translation to use it elsewhere. This requires selectable text. An image-only scan may need text recognition first.

This article is about

• Using OCR to translate scanned documents without retyping the text.
• Translating scanned PDFs and standalone JPG or PNG images with Lara Translate.
• Preparing readable source files and selecting the appropriate translation settings.
• Reviewing translated text, numbers, tables, and formatting before sharing.
• Troubleshooting missing text, recognition errors, and layout changes.

Sources

• Lara Translate: How to Translate Documents
• Lara Translate: Translating Images Inside PDF Documents
• Lara Translate: Image-to-Image Translation
• Lara Translate: What Is OCR?
• Lara Translate: AI + Human Translation
• Lara Translate: Character Pool, Extra Usage and Additional Costs
• Microsoft Support: Translate Text into a Different Language
• Apple Support: Interact With Text in a Photo Using Live Text in Preview on Mac
• Google Translate Help: Translate Documents and Websites
• Google Translate Help: Translate Images

L'articolo How to Translate Scanned Documents and Images proviene da Blog | Lara Translate.

A lot has changed since our first MCP explainer. ChatGPT, Gemini, and Microsoft Copilot now support the protocol, Anthropic handed it to a Linux Foundation project, and the July 2026 specification made it stateless. Here’s what MCP is today, how it works, where it’s still risky, and how Lara Translate plugs into it.

What is the Model Context Protocol, in plain terms?

MCP is a shared language between AI applications and the systems they need to reach. Large language models are good at reasoning over text, but on their own they can’t see your files, query your database, or post to your team’s channel. MCP gives them a standard way to ask.

Before MCP, each AI product shipped its own plugin or function-calling format. Connect five tools to three assistants and you were maintaining fifteen integrations. With MCP, the tool owner publishes one server, and every compatible assistant can use it. People often compare it to USB-C: one port, many devices.

MCP covers the connection only. It doesn’t dictate which model you use or how the app manages context once the data arrives, which is exactly why it travels so well between vendors.

How MCP has changed since 2024

If you read about MCP in early 2025, a good part of what you learned is now out of date. The protocol has gone through several specification revisions, picked up its biggest rival as an adopter, and moved to neutral governance.

The governance move is the one decision-makers should notice. The Agentic AI Foundation is a directed fund under the Linux Foundation, co-founded by Anthropic, Block, and OpenAI, with support from Google, Microsoft, AWS, Cloudflare, and Bloomberg. No single vendor controls the standard anymore. That lowers the lock-in risk of building on it.

How MCP works: hosts, clients, and servers

MCP uses a client-server model with three participants. Once you see how they fit, the rest of the protocol makes sense.

• MCP host: the AI application the user works in, such as Claude Desktop, Claude Code, ChatGPT, or VS Code.
• MCP client: a component the host creates for each server it connects to. Each client keeps one dedicated connection.
• MCP server: the program that provides context or actions, for example a filesystem, a database, or a SaaS product like Sentry.

Servers can run in two places. A local server runs on your machine and talks to the host over standard input/output (stdio). A remote server runs on the provider’s infrastructure and uses Streamable HTTP, typically with OAuth for authentication. Remote servers are why you can now connect a tool by pasting a URL instead of installing anything.

Under the hood, every message is JSON-RPC 2.0. You’ll rarely touch it directly, since official SDKs handle the plumbing in TypeScript, Python, C#, Go, Rust, Ruby, Java, Swift, PHP, and Kotlin.

What servers expose: tools, resources, and prompts

Servers offer three core building blocks, called primitives:

• Tools are actions the model can call: run a query, send a message, translate a paragraph.
• Resources supply read-only context, like file contents, database records, or a schema.
• Reusable templates that structure a task are prompts, such as a few-shot example set for a reporting job.

A database server might expose all three: a tool to run queries, a resource holding the schema, and a prompt with sample questions. On the other side, clients can support elicitation, which lets a server ask the user for missing input or confirmation before acting.

Stateless by design, and what that means for “memory”

Here’s a myth worth retiring. Early explainers (ours included) said MCP gives AI agents memory across sessions. It doesn’t, on its own. Since the 2026-07-28 revision, MCP is explicitly stateless: every request carries its protocol version and client capabilities, and servers advertise what they support through a server/discover call.

Memory, when you get it, comes from what a server stores. A server that needs state across calls mints an explicit handle and receives it back as a tool argument. The Lara Translate MCP server is a good example: translations you approve get saved to a translation memory, so the next request benefits from them. The protocol carries the call. The server keeps the knowledge.

Where MCP is used today

MCP started in code editors. It has since spread to almost every place people use AI at work.

Developer tools. Coding assistants were the first big adopters. The official MCP documentation uses VS Code as its reference example: when VS Code connects to the Sentry MCP server, it spins up a client that keeps that connection open, then adds another client for each additional server. Cursor, Claude Code, and OpenAI Codex follow the same pattern.

Data and analytics. Analysts no longer need to write SQL for every question. For example, the Kleene.ai MCP server lets teams ask Claude or ChatGPT which customer segment grew the most over the past 90 days, using data from their warehouse.

Everyday business apps. In September 2025, OpenAI brought full MCP support into ChatGPT’s developer mode, letting developers connect their own servers and tools directly to their accounts. Anthropic’s own count of MCP adopters includes ChatGPT, Cursor, Gemini, Microsoft Copilot, and Visual Studio Code.

Interactive interfaces. With the official MCP Apps extension, a server can render interactive elements such as charts, forms, or video players right inside the conversation. Servers still return plain text for hosts that don’t support it.

Translation and localization. Multilingual work is a natural fit, because a general-purpose model and a dedicated translation engine are good at different things. More on that below.

Is MCP secure? The risks to plan for

MCP gives AI real access to real systems. That’s the point, and it’s also the risk. The official MCP security best practices spell out several attack patterns that any team rolling out servers should know.

• Local server compromise. A local server runs with the same privileges as the client. A malicious package or a tampered startup command can read files or exfiltrate keys.
• Token passthrough. If a server forwards tokens that weren’t issued for it, attackers can bypass rate limits and audit trails. The specification explicitly forbids this.
• Watch for over-broad scopes, too: a stolen token with wildcard permissions opens every connected tool at once.
• Proxy servers face the confused deputy problem, where a reused consent cookie can let an attacker skip the user’s approval screen.

The fixes are mostly discipline. Install servers only from publishers you trust, request the smallest scopes that do the job, sandbox local servers where you can, and keep human approval on anything that writes, sends, or deletes. The official MCP Registry helps with discovery and verified namespaces, but a listing tells you where a server lives. It doesn’t certify that the server is safe.

How Lara Translate uses MCP

General-purpose LLMs handle many languages, but most of their training data is English. Domain terms, regional phrasing, and brand voice are where they slip. Lara Translate’s developer documentation describes the fix: let the LLM hand translation off to Lara’s Translation Language Models (T-LMs), which are trained on professionally translated text, and keep the LLM focused on reasoning.

There’s a cost angle as well. Tokenization and pricing are optimized for English, so translating with Lara Translate first means the LLM mostly processes English content. For teams running multilingual workflows at volume, that adds up.

What the Lara Translate MCP server can do

The Lara Translate MCP server exposes 21 tools in four groups:

• Translation: translate, with support for context, custom instructions, and automatic source language detection.
• Detecting languages takes two tools, detect_language and list_languages.
• Translation memories: eight tools to create, update, and delete memories, add or remove single translation units, and import TMX files.
• Managing glossaries is the biggest group, with eleven tools to create glossaries, add or delete entries, import CSV files, and export or count terms.

A quick example: context changes the answer

Ask an assistant to “translate with Lara ‘la terra è rossa’, I’m talking with a tennis player.” The assistant calls translate with the text, the target language, and the context “Conversation with a tennis player.” Lara Translate returns “The clay is red,” because in tennis terra means the clay court, not the ground.

Approve the result, and the assistant can call add_translation to save it to a named translation memory, then add_glossary_entry to lock the term for every future request. That’s how consistency builds up across a project, without anyone re-explaining the same terminology.

How to connect it

The fastest route is the hosted endpoint, https://mcp-v2.laratranslate.com/v1. Clients that support remote MCP servers and OAuth connect to it directly: you add the URL, log in with your Lara Translate account in the browser, and no API key goes into the configuration. In Claude, add it as a custom connector under Settings > Connectors. In OpenAI Codex, run codex mcp add and then codex mcp login.

If your client needs a local process, run the server with npx or Docker using an Access Key ID and Secret from your Lara Translate account:

{
  "mcpServers": {
    "lara-translate": {
      "command": "npx",
      "args": ["-y", "@translated/lara-mcp@latest"],
      "env": {
        "LARA_ACCESS_KEY_ID": "<YOUR_ACCESS_KEY_ID>",
        "LARA_ACCESS_KEY_SECRET": "<YOUR_ACCESS_KEY_SECRET>"
      }
    }
  }
}

Useful links to get started:

• Lara Translate MCP Server: setup and installation
• Why use Lara Translate inside an LLM
• Getting started with the Lara Translate MCP Server (Knowledge Base)
• Lara Translate MCP Server on GitHub
• Lara Translate MCP Server on Docker Hub

The bottom line

MCP grew from an Anthropic side project into shared infrastructure in under two years. It now has neutral governance, a stateless core, and support from the AI products most teams already use. The open questions have moved from “will it last?” to “which servers do we trust, and with what permissions?” Start small, connect one or two servers that remove a real bottleneck, and let specialized tools like Lara Translate handle the jobs a general model does only passably.

FAQ

Who created MCP, and who owns it now?

Anthropic created the Model Context Protocol and open-sourced it in November 2024. In December 2025, Anthropic donated MCP to the Agentic AI Foundation, a directed fund under the Linux Foundation. The foundation was co-founded by Anthropic, Block, and OpenAI, with support from Google, Microsoft, AWS, Cloudflare, and Bloomberg. MCP’s maintainers kept the same open, community-driven governance model after the move. In practice, no single AI company controls the standard.

Does ChatGPT support MCP?

Yes, ChatGPT supports MCP natively. OpenAI announced MCP support in March 2025, starting with its Agents SDK and extending to the ChatGPT desktop app and the Responses API. In September 2025, ChatGPT’s developer mode added the ability to connect your own MCP servers. You no longer need community bridges to use MCP with OpenAI models. The same server can serve Claude, ChatGPT, and other hosts at once.

How is MCP different from an API?

MCP is a standard layer that sits on top of APIs, so AI apps can use them in one consistent way. A traditional API is specific to one service, with its own authentication, endpoints, and error handling. An MCP server wraps that service and describes its tools, resources, and prompts in a format any MCP host understands. The model can then discover what’s available and call it without custom code for each pairing. Most MCP servers still call regular APIs behind the scenes.

Does MCP give AI agents memory?

No, MCP by itself does not store memory. Since the 2026-07-28 specification, the protocol is stateless, and each request carries everything the server needs to process it. Memory comes from the servers and apps you connect: a server can save data and hand back an explicit handle for later calls. The Lara Translate MCP server works this way, storing approved translations in translation memories you control. That keeps memory where you can inspect, edit, and delete it.

Is MCP safe to use at work?

MCP can be used safely at work, but safety depends on the servers you install and the permissions you grant. The protocol supports OAuth-based authorization, scoped permissions, and user consent prompts. The main risks are untrusted local servers, tokens passed through to other services, and overly broad scopes. The official MCP security best practices recommend least-privilege scopes, sandboxing local servers, and explicit consent before running new server commands. Treat each server like any third-party software you’d vet before installing.

How do I connect Lara Translate to Claude or other AI assistants?

Add the hosted Lara Translate MCP endpoint, https://mcp-v2.laratranslate.com/v1, to any client that supports remote MCP servers and OAuth. In Claude, go to Settings > Connectors, add a custom connector with that URL, and log in with your Lara Translate account. For OpenAI Codex, run codex mcp add followed by codex mcp login. Clients that need a local process can run the server with npx or Docker and your access keys. To test it, ask your assistant to “translate with Lara ‘Hello world’ to Spanish.”

This article is about

• What the Model Context Protocol is and the integration problem it solves
• How MCP evolved from Anthropic’s 2024 release to Linux Foundation governance and the stateless 2026-07-28 specification
• The roles of MCP hosts, clients, and servers, and the tools, resources, and prompts servers expose
• Why memory lives in MCP servers rather than in the protocol itself
• Real uses in coding, analytics, and business apps, plus the security risks to manage
• How the Lara Translate MCP server adds context-aware translation, memories, and glossaries to AI assistants

Sources

• Anthropic: Introducing the Model Context Protocol
• Anthropic: Donating the Model Context Protocol and establishing the Agentic AI Foundation
• SiliconANGLE: OpenAI adds support for Anthropic’s MCP
• VentureBeat: OpenAI adds MCP support in ChatGPT developer mode
• MCP Specification 2025-11-25: Key changes
• MCP Specification 2026-07-28: Key changes
• MCP Documentation: Architecture overview
• MCP Documentation: Official SDKs
• MCP Documentation: Extensions overview
• MCP Documentation: Security best practices
• MCP Blog: Introducing the MCP Registry
• Lara Translate Developer Docs: Why use Lara inside an LLM

This article was produced by the Lara Translate content team. Lara Translate is an AI translation platform built by Translated, with more than 25 years of professional translation experience. Teams use Lara Translate inside AI assistants and workflows through MCP to translate with context, glossaries, and translation memories, across 200+ languages and 60+ file formats.

L'articolo What Is the Model Context Protocol (MCP)? The Guide proviene da Blog | Lara Translate.

Lara 3 became generally available, while Lara Think added more control over reasoning and support for visual context. SOC 2 Type II is another headline for organizations evaluating Lara Translate for business use.

The integrations announced last month also moved forward: Lara Translate is now publicly available for Slack and published in the Make Marketplace. Audio Translation arrived on iOS and Android, and TBX glossary import and export reached the web app.

Here are the Lara Translate October 2026 updates: what shipped, and why it matters.

Lara 3 is now generally available

When you translate professionally, a fluent sentence is only the starting point. The terminology, meaning, and tone also need to survive the move into another language.

Lara 3, the third generation of Translated’s translation-specialized AI model, became generally available in October through Lara Translate, the Lara API, and TranslationOS.

In Translated’s blind evaluation of 16,800 translations from English into 21 enterprise languages, professional translators judged around 82% of Lara 3 translations accurate without edits. Lara 3 ranked first in 20 of the 21 language pairs tested.

For translators and localization teams, the practical question is how much work remains after the first translation. These results provide a useful starting point for testing Lara 3 against your own content and review standards.

Read the Lara 3 announcement and evaluation results.

Lara Think: more control over reasoning, with visual context

Some translation decisions need a closer look. An expression may carry a cultural meaning, or a short product label may be impossible to interpret without seeing what it describes.

Lara Think analyzes the source before producing and revising a translation. October adds selectable reasoning effort levels, allowing developers to balance deeper analysis with processing time.

Lara Think text translation also accepts an accompanying image or image URL through the API and supported SDKs. A screenshot, packaging image, or product photograph can provide context that the words alone leave out. Image context is currently exclusive to Lara Think.

There is benchmark news, too. As of October 20, 2026, Lara 3 Think ranked first on the Last Translation Benchmark’s Blind leaderboard, with a 56.09% pass rate on its 911-item evaluation set. The test focuses on difficult examples that leading translation systems had failed, including slang, idioms, and cultural references.

Read more about Lara Think and the benchmark.

SOC 2 Type II: supporting enterprise security reviews

Translation quality is one part of choosing a service for business use. Security teams also need evidence about how the provider manages and protects the information entrusted to it.

SOC 2 Type II is another major milestone this month. For organizations evaluating Lara Translate, it supports the due diligence process that accompanies the adoption of a new translation service.

The Translated Trust Center is the starting point for reviewing published security information and requesting access to supporting documentation. Procurement, IT, and security teams can use it to assess the relevant evidence and clarify how it applies to their intended use of Lara Translate.

Explore the Translated Trust Center.

Lara Translate reaches Slack and Make

September’s article introduced Slack and Make as integrations going through review. October brings both into users’ workflows.

Translate within Slack conversations

Lara Translate for Slack is now publicly available. Multilingual teams can access Lara within Slack, reducing the need to move content between a workplace conversation and a separate translation tool.

This is particularly useful when translation is part of an ongoing exchange: understanding a colleague’s message, preparing a reply, or sharing information across teams working in different languages.

The app is available for public distribution. A Slack Marketplace listing is still pending, so public availability and Marketplace listing remain separate milestones.

Add translation to Make automations

Lara for Make is approved and published in the Make Marketplace. You can now incorporate Lara translation capabilities into automation scenarios connecting your applications and business processes.

For teams with repeatable translation tasks, this provides a way to make translation one step in a broader workflow, without building a custom integration from scratch.

Explore Lara Translate in the Make Marketplace.

Audio Translation arrives on iOS and Android

A voice message or interview recording should be as easy to translate as a paragraph of text.

Audio Translation is now available on both iOS and Android. You can record directly in Lara, choose an existing audio file, or send audio from another app through your phone’s sharing options. Once translated, the audio can be played and shared from the mobile experience.

This makes the feature useful for voice messages, interviews, and other recordings where the voice is part of the experience.

Each app also makes use of its platform’s native controls. On Android, recording supports press-and-hold capture, hands-free operation, and background recording with a persistent notification. You can pause, preview, and resume a recording.

On iOS, Live Activities and Dynamic Island let you follow recording, upload, and translation progress after leaving the app.

Get Lara Translate on your phone

Download the app and keep translation ready when you need it.

Download for Android | Download for iOS

TBX glossary import and export, now in the web app

If your team already maintains a termbase, bringing that terminology into Lara should fit the way you work.

You can now import and export glossaries in TBX format directly through the Lara Translate web app. TBX, short for TermBase eXchange, is a format used to exchange terminology databases between translation and localization tools.

September introduced TBX support through the API and SDK. October brings the same capability to people managing glossaries through the web interface.

For translators and terminology managers, this removes the need to convert a TBX termbase into CSV before importing it into Lara. You can also export Lara glossaries in TBX format for use elsewhere in your localization workflow.

For developers, existing API and SDK support continues to provide a programmatic route for exchanging terminology. Teams can work through the web app or their own integrations using the same exchange format.

Explore the Lara Translate developer documentation.

Try October’s updates with your own content

Start with a translation you know well: a product description with specific terminology, a passage with an ambiguous expression, or a voice message you need to share in another language.

Use it to assess Lara 3, explore Lara Think’s reasoning capabilities, or try the new mobile audio workflow. If translation is a recurring part of your team’s work, Slack, Make, and TBX glossary exchange give you more ways to connect Lara to that process.

Try Lara Translate.

FAQ

Are Slack and Make both available?

Yes. Lara Translate for Slack is publicly available, although its Slack Marketplace listing is still pending. Lara for Make is published in the Make Marketplace.

What changed for Lara Think in October?

Developers can choose reasoning effort levels and provide an image or image URL as context for text translation through the API and supported SDKs.

Can I translate audio on my phone?

Yes. Both the iOS and Android apps support recording or importing audio, translating it, and playing or sharing the result.

Is TBX support new this month?

TBX import and export is new in the web app. Support through the API and SDK was introduced previously.

This article is about

• Lara 3’s general availability
• Lara Think reasoning controls, image context, and benchmark results
• SOC 2 Type II and enterprise security reviews
• Lara Translate integrations for Slack and Make
• Audio Translation on iOS and Android
• TBX glossary import and export in the web app

L'articolo Lara Translate October 2026 Updates: Lara 3, Think, and more proviene da Blog | Lara Translate.

Connecting WordPress, n8n, and Lara Translate lets you automate much of that process. WordPress supplies the article, n8n passes the content between tools, and Lara Translate translates it. The automated translation workflow creates drafts for an editor to review before publishing.

We tested this workflow on n8n Cloud with an English WordPress article and Italian as the target language. It retrieved the post, translated the title and HTML body, and created a separate WordPress draft.

What is an Automated Blog Translation Pipeline?

An automated blog translation pipeline retrieves source content, sends it for translation, and saves the result for review. You configure the steps once and reuse them for later posts.

You can also automate WordPress translation through Claude using Lara Translate’s MCP server. Our guide on how to automate blog translation with Claude, Lara Translate, and WordPress explains that setup, including saving translated posts as drafts for human review.

What Do You Need to Automate WordPress Translation?

Before building the automated WordPress translation workflow with n8n and Lara Translate, make sure you have:

• A WordPress account that can read the original posts and create drafts through the REST API, the interface external tools use to work with your site.
• An n8n instance where you can install the Lara Translate verified community node.
• A Lara Translate account with API credentials and an available usage allowance or billing arrangement.
• One source article and one target language for the initial setup.
• A multilingual plugin or equivalent site setup if readers need linked language versions.

How Do You Connect WordPress and Lara Translate to n8n?

You will build and manage the complete automated translation workflow inside n8n. First, connect your WordPress site so n8n can retrieve articles and create drafts. Then connect Lara Translate so n8n can send content for translation and receive the result.

You do not need to set up a separate connection between Lara Translate and WordPress for this workflow because n8n passes the content between them.

Step 1: Connect Your WordPress Site to n8n

On a self-hosted WordPress site with Application Passwords enabled, open your user profile and find Application Passwords. Create one with a recognizable name, such as “n8n blog translation.” Click Add Application Password and copy the generated password before leaving the page.

The application password has your WordPress account’s permissions and can be revoked separately from your login password. If the option is missing, ask your site administrator to check whether the feature is enabled. Follow WordPress’s application-password guide for the setup.

In n8n, create a WordPress credential, select Basic Auth, and enter:

• Your WordPress site URL.
• Your WordPress username.
• The application password you just created.

Click Save. Reuse this credential in both WordPress nodes. Keep Ignore SSL Issues switched off for a site with a valid HTTPS certificate.

If your site is hosted on WordPress.com, follow the applicable authentication instructions in the n8n credentials guide.

Step 2: Connect Lara Translate to n8n

In n8n Cloud, open the workflow’s node picker, search for Lara Translate, and install the verified community node.

For a supported self-hosted instance, follow n8n’s community-node installation instructions and use the package n8n-nodes-lara-translate.

Create your Access Key ID and Access Key Secret using Lara Translate’s API key guide. In n8n, add a Lara Translate node and create its credentials. Enter both keys and click Save. Reuse this credential for title and document translation.

Once both credentials are configured, you can build the steps that retrieve, translate, and save your article. Lara Translate’s n8n integration guide explains the supported operations and credential setup.

How Do You Build the Automated WordPress Translation Workflow?

To build the automated WordPress translation workflow, start with one article and one target language. Choose a post with the headings, links, images, and any page-builder blocks you normally publish. Keep the first run manual. This gives you a chance to inspect the content at each stage before allowing the workflow to process posts on a schedule.

Step 3: Retrieve the Article You Want to Translate

Create a new workflow and add a Manual Trigger, followed by a WordPress node. Select your saved credential, Resource: Post, and Operation: Get. Enter your source post ID.

Execute the node and inspect title.rendered and content.rendered. WordPress can include HTML entities in title.rendered, such as ’ for an apostrophe or & for an ampersand. Check the source title before translation and review the translated title for any entity codes that remain visible.

Step 4: Prepare the Article’s Fields

Add Edit Fields (Set) directly after Get a post. Select Manual Mapping and create the five fields below. Keep field names in Fixed mode and use Expression mode for the values containing {{ … }}.

This workflow uses content.rendered, which contains the article’s rendered HTML rather than its original Gutenberg block structure. The translated draft will usually open as a single Classic block. Keeping HTML tags and classes does not preserve the original editable blocks.

The source post ID and modification time remain available in Edit Fields, but the translation branches do not pass them into Merge. This example processes one article at a time.

Step 5: Translate the Title and HTML Body

Create two branches from Edit Fields. Connect one to Lara Translate → Translate Text and the other to Convert to File.

In Translate Text, select your Lara credential, Resource: Translation, Operation: Translate Text, Source Language: English, and Target Language: Italian. Set Text to Translate to the expression {{ $json.source_title }}. Under Additional Options For Translate Text, add Content Type and select Plain Text. Execute the title node.

The node also exposes glossary, translation memory, instructions, and style options. You can use your team’s approved resources when configuring terminology controls.

Translate the Body as HTML

The Translate Text operation in the n8n node we tested accepts Plain Text and XLIFF but not HTML. Use Translate Document for the article body, so Lara Translate can process it as an HTML file.

Document translation costs: Paid API document translation has a minimum charge of 20,000 characters per document, even when the file contains fewer characters. Short blog posts can still be translated, but each document request is billed for at least 20,000 characters.

1. On the second branch from Edit Fields, add Convert to File. Keep the title node on its separate branch: its output does not contain source_body_html.
2. Select Convert to Text File. Set Text Input Field to the fixed field name source_body_html. Set Put Output File in Field to data. Under Options, set Encoding to utf8 and File Name to source.html. Execute the node.

3. Connect Lara Translate → Translate Document after Convert to File. Select your Lara credential, Resource: Translation, Operation: Translate Document, Source Language: English, and Target Language: Italian. Set Input Binary Field to data and Document Name to source.html.
4. Leave Output Format unset and execute the node.

5. Add Extract from File after Translate Document and choose Extract From Text File. Set Input Binary Field to data, Destination Output Field to translated_body_html, and File Encoding to utf8. Execute this node.
6. The output will contain Italian paragraphs in translated_body_html. The visible HTML retains paragraph tags and wp-block-paragraph classes.

This workflow sends the existing body HTML without adding an outer document wrapper. If your setup adds <html>, <head>, or <body> elements, use only the article content inside <body> when creating the WordPress draft.

Keep these branches separate until Merge. Connecting Translate Text after Extract from File makes {{ $json.source_title }} empty because that input only contains the extracted body.

Step 6: Merge the Results and Create a WordPress Draft

Add a Merge node. Connect Translate Text to Input 1 and Extract from File to Input 2. Select Mode: Combine, Combine By: Position, and Number of Inputs: 2. Run the workflow and inspect Merge.

The output should contain one item with translation and translated_body_html. Position combines items by their order.

Add a WordPress node directly after Merge. Select your saved WordPress credential, Resource: Post, and Operation: Create.

Map the fields below using Expression mode for Title and Content. Under Additional Fields, add Content and Status, then choose Draft.

Click Execute step once. Operation: Create makes a separate post. Repeating this step will create another draft.

WordPress lists the available fields in its Posts REST API reference.

Optional: Assign the Language and Link the Translation

Creating an Italian draft does not automatically tell your website which English article it belongs to. Configure that relationship through the multilingual system your site already uses.

For example, Polylang Pro’s REST API supports assigning a language with lang and linking posts through translations. Those fields require Polylang Pro and are not standard WordPress parameters.

If your site uses WPML or another system, follow its integration method. You may need an authenticated HTTP Request node where the built-in WordPress node does not expose the required operation.

Match the translation language to the language configured on your site. Also check localized URLs, internal links, SEO titles, meta descriptions, and hreflang output.

Step 7: Review the Draft in WordPress

Open Posts → All Posts → Drafts and find the translated title. In the image below, the editor shows “10 luoghi dove cercare lavori freelance” with Italian body text and Draft status. This confirms that retrieval, translation, and draft creation worked together.

Compare the translation with the original, then preview the page to check headings, links, lists, and spacing. Because this workflow sends rendered HTML back to WordPress, the body will usually appear as a single Classic block. Select the Classic block and choose Convert to blocks if you want separate editable blocks, then review the result again. Conversion may not recreate custom blocks or page-builder layouts.

Images containing English text also need attention. Decide whether to replace them, translate them separately, or explain the information in a caption.

Publish after the editor has checked the translation, layout, and language settings.

Step 8: Add Scheduling and Duplicate Checks (Optional)

Once the manual draft passes review, add a Schedule Trigger and set its frequency and timezone. Save and publish the workflow to enable scheduled execution. Publishing the n8n workflow enables the schedule. The WordPress post remains a draft because its creation status is set to Draft.

Replace the fixed post selection with rules for eligible articles. For example, retrieve published English posts in a selected category that still need an Italian translation. Exclude translated posts so the workflow does not pick up its own output.

Keep a translation log in n8n Data tables or a database. Record the original post ID, target language, source modification time, translated draft ID, and processing status.

Before translating any post into Italian or another language, check that combination in the log. If its draft already exists and the original has not changed, skip it. If the original has changed, flag it for your update process rather than overwriting an editor’s corrections.

Account for overlapping runs and partial failures, too. Make sure two runs cannot process the same job at once. If WordPress creates a draft but the log update fails, recover that draft before repeating the creation step.

How Do You Keep Terminology Consistent Across Translated Posts?

Suppose your product has a feature called “Workspace.” One article keeps that name in English, while the next uses an Italian translation. Both may read well, but readers now have two names for the same feature.

Lara Translate supports glossaries for preferred terms and translation memories for previous translations. You can import a translation memory as a TMX file through its interface. Select the relevant resources in your workflow rather than assuming a choice in the web app applies to every API request.

Give the reviewer your approved product names and phrases, so they can check the article against the same reference.

When an editor corrects a phrase in WordPress, decide how that approved change will reach your translation memory. Editing the post alone does not create that connection. A separate update process lets you reuse approved wording without treating every unreviewed draft as reference material.

What Should You Check If The Workflow Fails?

If the WordPress translation pipeline fails, open the execution and inspect the node that failed before starting again. A translation request that stopped and a WordPress request that created a draft need different next steps.

Conclusion

Once the automated WordPress translation workflow is configured, a new article can move from WordPress through translation and return as a draft for review. The editor can focus on the actual language and page layout, with fewer fields to copy between tools.

Lara Translate translates the title and article body within your n8n workflow, so you can create a WordPress draft without copying content between tools. Start with one post and review the result. Before scheduling regular runs, add checks to avoid duplicate drafts and protect an editor’s changes when the original article is updated.

Frequently Asked Questions

Can you automate WordPress blog translation with n8n?

Yes, you can automate WordPress translation with n8n. n8n supports retrieving and creating WordPress posts, and Lara Translate supplies a translation node. Connect those operations, map the content fields, and configure scheduling and tracking for ongoing processing.

Do you need coding skills to connect Lara Translate to n8n?

The verified community node supports a visual setup for credentials, languages, and input mapping. HTML preparation, custom WordPress fields, and plugin-specific linking can require technical help, depending on the site.

Does a paid Lara Translate plan include API usage?

No. API usage, including the n8n integration, is billed separately from the character quota in Pro and Team plans. A free account is enough to start. Check your API allowance before turning on scheduled runs.

Can Lara Translate translate HTML blog content?

Yes. Lara Translate supports HTML files through the Translate Document node in n8n. You can convert the blog body’s HTML into a file, translate it, and extract the translated HTML to create a WordPress draft. Review the page’s formatting before publishing.

Do you need WPML or Polylang for this workflow?

You can create a translated draft without either plugin. A multilingual site still needs a way to assign languages and connect translations. You should follow the site’s existing plugin or equivalent architecture.

Can translated posts be saved as drafts instead of published automatically?

Yes, translated posts can be saved as drafts instead of being published automatically. You should set the WordPress creation status to draft in the workflow, so an editor can review and publish it later.

This article is about

• Connecting WordPress, n8n, and Lara Translate for automated translation.
• Translating post fields and handling HTML content.
• Reviewing drafts and keeping terminology consistent.
• Scheduling translations, avoiding duplicates, and linking language versions.

Sources

• Lara Translate: n8n integration overview
• Lara Translate: official n8n setup guide
• Lara Translate: n8n integration reference
• Lara Translate: Translate Text SDK reference
• Lara Translate: API credentials guide
• Lara Translate: glossaries
• Lara Translate: translation memories
• Lara API: Document Translation 
• n8n: WordPress node reference
• n8n: community-node installation
• n8n: Convert to File
• n8n: Extract From File
• n8n: Schedule Trigger
• n8n: Data tables
• n8n: Merge
• WordPress: application passwords
• WordPress Classic block
• n8n WordPress credentials
• WordPress: Posts REST API
• Polylang Pro: REST API
• Lara Translate Pricing
• Lara Translate Character Pool: Extra Usage and Additional Costs

L'articolo How to Build an Automated WordPress Translation Pipeline (n8n + Lara Translate) proviene da Blog | Lara Translate.

Errors like this slip through because the output looks fluent. The Last Translation Benchmark was built to collect exactly these cases. As of 20 October 2026, Lara 3 Think, the reasoning model in Lara Translate, tops its leaderboard with a 56.09% pass rate. Here’s how the benchmark works, how the scores compare, and when reasoning is worth the extra seconds.

What is the Last Translation Benchmark?

The Last Translation Benchmark (LTB) is an academic project led by Vilém Zouhar, with ETH Zurich among its core institutions. It brings together 260 contributors from 177 institutions. Its first release, LTBv1, contains 3,456 human-written, peer-reviewed examples across 109 languages, covering text, images, audio, and video.

The selection rule is strict. An example was accepted only if most leading machine translation systems got it wrong. Why so harsh? The authors argue that existing translation benchmarks are close to saturation, so a harder test was needed to separate the best systems.

The result leans heavily on informal and culture-bound language. Across the full dataset, the paper’s labels include 986 cultural artifacts and 453 slang items. The leaderboard scores models on LTBv1-eval, a 911-item, text-only subset.

How the Last Translation Benchmark is scored

Every example comes with pass/fail rules written by the contributor who submitted it. A rule might say that “Pfund” has to become 500 g, or that a proverb has to keep its figurative meaning. A translation passes only if it meets every rule.

An LLM checks each translation against those rules. The live leaderboard currently uses Gemini 3.8 Flash as the verifier, while the paper’s interactive verification used Gemini 3.1 Pro. The paper reports a Kendall τb of 90.5 (on a 0 to 100 scale) between the verifier’s system rankings and human rankings. It also recommends open-source LLMs as verifiers for the highest reproducibility.

Funding is disclosed, too. The paper thanks Cohere, Google, and OpenAI for financial support to run the platform at scale, and Charles University for server infrastructure.

The leaderboard has two modes. Every figure in this article comes from the main Blind mode, where models translate without seeing the rules or the human reference.

Last Translation Benchmark results (October 2026)

Top AI systems on LTBv1-eval, Blind mode, as of 20 October 2026. Source: LTB leaderboard.

Human translations by the benchmark contributors score 99.89%, which sets the ceiling.

Lara 3 Think leads the next system by almost 11 percentage points. That’s roughly 98 more sentences translated correctly out of 911, on a test where every item was picked because leading systems failed it. Lara 3 Think ran with the Fluid style and no adaptation at all: no context, translation memories, glossaries or custom instructions.

A quick note on names. Lara Think is the reasoning mode in Lara Translate. Lara 3 Think is its current version, built on Lara 3, and Lara 2 Think Pro belongs to the previous generation.

Why do Lara Base and Lara Think score so differently?

The same leaderboard lists an earlier standard Lara model (a November 2025 entry) at 0.66%. That figure says more about how the run went than about the quality of the translations. Our logs from those benchmark runs show a long list of technical errors, caused mainly by differences between the standard input formats Lara Translate expects and the formats used to run the benchmark. The public dataset reflects this: 200 of the 911 evaluation items have no Lara output at all, and every one of them counts as a fail.

We’re working with the benchmark team to fix the integration and run the evaluation again, so the leaderboard reflects what Lara Translate actually delivers.

What is Lara Think and how does it work?

Lara Think is the reasoning model in Lara Translate. Before it commits to a translation, it runs a multi-step linguistic analysis of the source, takes glossaries, style guides, and context into account when they’re available, and then revises its own output.

Lara 3 also introduced a training phase. The model generates alternative translations, an automatic judge scores them, and the model refines itself over millions of cycles. According to Translated, Lara Think “cuts errors nearly in half when higher quality is required.”

The trade-off is time. Lara Think can take up to 10x longer than Lara Base. For developers, a callback can return a fast preliminary translation first while the reasoned version completes.

What kinds of errors does the benchmark catch?

These items come from the LTBv1 evaluation set. Each one shows what the source means and what a frontier model produced.

Examples from LTBv1-eval with Gemini 3.1 Pro output.

None of these outputs is ungrammatical. Each one passes a quick read, and each one changes the meaning. In professional settings, that’s the most expensive kind of error: nobody flags it until a customer, a reader, or a court does.

Who is Lara Think for?

Pick Lara Think when a wrong nuance costs more than a few extra seconds:

• Marketing and brand copy, where tone and wordplay carry the message
• If you handle legal, financial or regulated content, one term can change an obligation
• Publishing and editorial work with dialogue and informal registers
• Reviews, social posts and other user-generated content full of slang
• Building a quality-critical localization pipeline? This is the model for it.

For real-time chat, very high volumes, or latency-sensitive workflows, Lara Base is the better fit. Speed matters more there than the last few hard cases.

How to try Lara Think

Developers enable Lara Think through the Lara Translate API by passing reasoning: true in the options object. The developer documentation covers SDKs, callbacks and pricing.

The bottom line

Benchmarks built from easy sentences can no longer tell strong translation systems apart. The Last Translation Benchmark can, because every item is a trap. On that test, Lara 3 Think leads by almost 11 points, and the reasoning step is the reason. Use it where a missed nuance is expensive, and keep Lara Base for speed.

FAQ

What is the Last Translation Benchmark?

The Last Translation Benchmark is an academic test of AI translation built from human-written examples that most leading machine translation systems get wrong. Its first release, LTBv1, has 3,456 examples across 109 languages, submitted by 260 contributors from 177 institutions.

Which AI translator scored highest on the Last Translation Benchmark?

Lara 3 Think, the reasoning model in Lara Translate, scored highest with a 56.09% pass rate on the 911-item LTBv1-eval set in Blind mode, as of 20 October 2026. GPT-6 Astra Pro is second at 45.33%.

How is the Last Translation Benchmark scored?

An LLM verifier checks each translation against pass/fail rules written by the contributor, and a translation passes only if it meets all of them. The live leaderboard uses Gemini 3.8 Flash as the verifier.

What is Lara Think?

Lara Think is the reasoning model in Lara Translate. It analyzes the source text in several steps before translating, which improves accuracy on ambiguous, informal and culture-bound content.

What’s the difference between Lara Think and Lara Base?

Lara Base is the default model, optimized for speed and consistency on professional content. Lara Think adds a reasoning layer for higher quality and can take up to 10x longer.

How much does Lara Think cost?

Lara Think costs €2,000 per million source characters on the Pro plan and €1,500 on the Team plan, through the Lara Translate API. It isn’t available on the Free plan.

This article is about

• Lara 3 Think ranks first on the Last Translation Benchmark with 56.09%, almost 11 points ahead of GPT-6 Astra Pro.
• How the benchmark is built: 3,456 examples that leading AI translators failed, from 260 contributors in 109 languages.
• Why an LLM verifier with contributor-written rules scores the results, and who funds the platform.
• Why reasoning, not raw translation training, decides performance on slang, idioms and cultural references.
• When to choose Lara Think over Lara Base, and what it costs through the API.

Sources

• Last Translation Benchmark: leaderboard
• Last Translation Benchmark paper (arXiv 2609.04173)
• LTBv1 dataset on Hugging Face
• Introducing Lara 3 (Translated)
• Translate Text: Lara Translate for Developers

L'articolo Last Translation Benchmark: Lara Think Ranks First proviene da Blog | Lara Translate.

Lara is the translation model behind Lara Translate, built by Translated. Its third generation, Lara 3, became generally available in October 2026, after a first release to selected partners in July. You can now use Lara 3 in Lara Translate, the Lara API, and TranslationOS. Here’s what the evaluation found, why it matters if you translate for a living or at scale, and where to try it.

How do you test a translation model fairly?

Grading your own homework is easy. It’s also not very convincing. So the evaluation put Lara 3’s output in front of professional translators, blind: they saw each translation without knowing whether it came from Lara 3, GPT-5.6 Sol, Claude Fable 5, DeepL, or Google Translate.

The test covered 16,800 translations from English into 21 enterprise languages. Part of the material came from WMT25, the 2025 edition of a standard public machine translation benchmark that spans books, news, and user conversations. The rest was proprietary enterprise content from travel, technology, and finance, which is the kind of text companies actually send out for localization.

How often did Lara 3 translations need no edits?

Translators care about one practical question: can I use this as it is? Around four in five Lara 3 translations were judged accurate, meaning they needed no edits at all. For the strongest general-purpose model in the test, the figure was roughly two in three.

Averaged across all 21 language pairs, the scores came out like this.

Average scores from blind professional evaluations of 16,800 translations, English into 21 enterprise languages. Source: Translated, October 2026.

Lara 3 ranked first in 20 of the 21 language pairs.

Where does Lara 3 make the biggest difference?

Enterprise localization is where most translation tools start to wobble. The approved term every single time, the same product name across hundreds of strings, a voice that sounds like the brand in every market: that’s a lot to hold at once. According to Translated, Lara 3’s advantage was most evident in exactly this kind of work, where terminology, context, consistency, and brand voice matter most.

This is the job Lara was built for. In Lara Translate, Lara works with your glossaries and translation memories, so the terms and phrasing your team has already approved carry over into every new translation.

What happens when you need to translate a lot, fast?

Localizing a product catalog, a help center, or a steady stream of support tickets turns speed and budget into hard limits. A model that translates well but slowly, or well but expensively, forces you to pick which content gets translated at all. Nobody likes that trade-off.

In a separate single-thread benchmark, Lara 3 delivered 23 times the throughput of Claude Fable 5. For the same budget, it translated 3.75 times as many characters. For teams working at volume, that means more content shipped per day and per euro.

How was Lara 3 trained?

Most translation models learn by imitating. They study translations people have already made and learn to do the same. The problem is that a sentence can often be translated correctly in more than one way, so a model that only copies can only be as good as the examples it learned from.

For Lara 3, Translated added one more step to the model’s training, and it works a lot like practice. Lara translates the same sentence in several different ways. A judge, built from the work of professional reviewers, scores each version and points out every mistake, down to a single word that doesn’t fit. Lara learns from the versions that scored well, corrects the mistakes, and starts again. Millions of times. You can read the full technical story in our introduction to Lara 3.

Why build a model only for translation?

Lara is built to do one thing: translate. Translated has been building language technology since 1999, and designing a model around a single purpose was a bet the company made early. The Lara 3 results show that a translation-specialized model can outperform the largest general-purpose AI models in a field where accuracy is non-negotiable.

Marco Trombetti, founder and CEO of Translated, put it this way:

Translation is a particularly unforgiving domain for AI as technical fluency alone is not enough. An ambiguous meaning, an incorrect term or a missed cultural cue can render a translation inaccurate and unusable. We’ve always believed that understanding these intricacies deeply and building a model specifically designed to grapple with them, is the only way to deliver genuinely quality translation. Lara 3’s performance is proof that our approach works. These results show what is possible when you commit to developing AI with one defining purpose. You don’t need to be the biggest player in AI to lead, you need to know exactly what problem you’re addressing and how to solve it.

Where can you use Lara 3?

If you translate on your own or as part of a team, you’ll find Lara 3 in Lara Translate, the online AI translator for text and documents in 200+ languages.

Enterprises and professional translators can also work with Lara 3 through the Lara API, which plugs the model into your own products and pipelines, and through TranslationOS, Translated’s AI service delivery platform. Building with the API? Start with the Lara Translate developer documentation.

The bottom line

Reading blind, professional translators preferred Lara 3 over the biggest general-purpose AI models and the best-known translation tools. It gets more translations right the first time, it’s faster, and it stretches the same budget further. The quickest way to judge it is to hand it something you actually need translated.

FAQ

What is Lara 3?

Lara 3 is the third generation of Lara, the translation-specialized AI model developed by Translated. It powers Lara Translate and is also available through the Lara API and TranslationOS.

When was Lara 3 released?

Lara 3 became generally available in October 2026. Translated first released it to selected partners in July 2026.

Is Lara 3 better than DeepL and Google Translate?

In blind evaluations by professional translators, yes: Lara 3 averaged 82% across 21 language pairs, against 57% for DeepL and 53% for Google Translate. It also scored ahead of GPT-5.6 Sol (69%) and Claude Fable 5 (67%), and ranked first in 20 of the 21 pairs.

How was Lara 3 evaluated?

Professional translators reviewed 16,800 translations from English into 21 enterprise languages without knowing which system produced each one. The content combined the WMT25 public benchmark (books, news, and user conversations) with proprietary enterprise material from the travel, technology, and finance sectors.

Is Lara 3 faster than general-purpose AI models?

Yes. In a single-thread benchmark, Lara 3 delivered 23 times the throughput of Claude Fable 5 and translated 3.75 times as many characters for the same budget.

How can I start using Lara 3?

Open Lara Translate, the online AI translator, to use Lara 3 for text and documents. Enterprises and professional translators can also access Lara 3 through the Lara API and TranslationOS.

This article is about

• Lara 3, Translated’s translation-specialized AI model, reaching general availability in October 2026
• A blind evaluation in which professional translators scored 16,800 translations into 21 enterprise languages
• Lara 3’s 82% average score against GPT-5.6 Sol, Claude Fable 5, DeepL, and Google Translate
• Why enterprise localization, with its terminology and brand voice demands, showed the widest gap
• Throughput and cost results against Claude Fable 5, plus the “learning by doing” training phase
• Using Lara 3 in Lara Translate, the Lara API, and TranslationOS

Sources

• Translated Announces General Availability of LARA 3, MultiLingual
• Introducing Lara 3, Translated
• Lara 3 Overview (benchmark methodology), Translated

This article was produced by the Lara Translate content team. Lara Translate is an AI translation platform built by Translated, with more than 25 years of professional translation experience. Individuals, teams, and enterprises use Lara Translate to translate text and documents with Lara 3, across 200+ languages and 70+ file formats.

L'articolo Lara 3 Is Now Available: The Translation Model Professional Translators Preferred in Blind Tests proviene da Blog | Lara Translate.

The headline is Lara Translate in ChatGPT: the plugin is now published, so you can translate without leaving the conversation. For teams, a new Group Admin role makes Lara Translate easier to run as an organization grows. My Translations reached its first production release across web, iOS, and Android. And there is more across mobile and the developer tools.

Here are the Lara Translate September 2026 updates: what shipped, and why it matters.

Lara Translate is now available in ChatGPT

A lot of drafting, research, and content work already happens inside ChatGPT. Until now, translating any of it meant leaving that flow, running the text through another tool, and bringing the result back.

The Lara Translate plugin for ChatGPT is now live. It maps Lara Translate’s existing MCP tools into the OpenAI environment, so you can use Lara Translate’s context-aware translation directly inside a ChatGPT conversation, as part of your research, writing, and content work.

You can find and install Lara Translate from the ChatGPT Plugins Directory. Once installed, you can invoke it directly in a conversation, including through an @ mention or from the + menu where available. Availability can depend on your plan and workspace settings.

This is the first of several. The Lara Translate integrations for Claude, Slack, and Make are currently going through their respective marketplace review processes.

Open Lara Translate in ChatGPT.

What’s new for teams: the Group Admin role

As a team grows, one person often ends up managing everything: who belongs where, which glossaries and translation memories each part of the business can use, and how usage breaks down by department. That becomes hard to sustain in larger, more distributed organizations.

Groups were introduced to address two recurring needs from larger customers: keeping linguistic resources separated, and reporting usage by organizational unit. With Groups, an organization can keep Translation Memories and Glossaries separated between different parts of the business, for example so a Legal team does not draw on resources meant for Marketing. Group-level usage reporting then gives clearer visibility into how Lara Translate is being consumed across teams or departments.

September adds the Group Admin role for Team accounts, including Enterprise accounts, which run on Team plans with additional needs. A Team Owner can now delegate the day-to-day management of an individual Group to a dedicated Group Admin, while keeping ownership of the overall Team account. The difference is one of scope: a Team Owner owns and controls the entire Team account, while a Group Admin manages a single Group on the Owner’s behalf. This makes Lara Translate easier to operate as responsibilities spread across a larger organization.

See how Groups and resource sharing work in Lara Translate.

My Translations, now across web and mobile

You translate a paragraph, refine it until it reads right, then move on and lose it. The next time the same text comes up, you start over.

My Translations reached its first production release, and it is available across the web app, iOS, and Android. You can save a text translation, including the source, the translated content, and any edits you made, then reopen it, edit it again, or remove it when you no longer need it. It gives you a persistent personal library of useful translation work to return to without repeating the original workflow.

On mobile, you can save directly from the Text view and open your collection from the bookmark entry point or from Account, then Resources. This first milestone focuses on the core save, browse, retrieve, edit, and delete experience across every platform.

Learn how My Translations works.

What’s new for image translation: Generative mode in the web app

When you translate an image, the new text rarely fits the old space. A short phrase runs long, a caption shifts, and the layout can end up looking patched together.

Generative Image Translation is now available directly in the Lara Translate web app. You can choose between two modes depending on the result you need. Inpainting replaces the original text while keeping most of the image unchanged. Generative mode recreates the image around the translated content, adapting typography, spacing, and layout for a more natural and visually consistent result.

This brings a capability previously available through Lara Translate’s developer ecosystem directly to web-app users, and gives you control over the trade-off between cost, visual fidelity, and how much of the original image is modified.

And it’s also available on our mobile apps!

Get Lara Translate on your phone

Save translations, use Meaning in Context, and sign in with SSO, all from your phone.

Download for Android | Download for iOS

What’s new on mobile: SSO on iOS and Meaning in Context

Two mobile updates stand out this month, one for organizations and one for anyone working with an unfamiliar text.

Single Sign-On is now on iOS. Eligible users can access Lara Translate with their organization’s company credentials. Alongside SSO, we improved sign-up and login reliability, including authentication flows that could unexpectedly end an active session.

Meaning in Context is now live on both iOS and Android. You can view the meaning of an individual word directly within the text you are translating, with the explanation generated from the surrounding context. It is most useful for ambiguous words or expressions whose meaning depends on the surrounding context, thereby extending Lara Translate toward contextual language understanding.

See how Meaning in Context works.

What’s new for developers: TBX glossaries and EPUB

Professional localization and publishing teams work with formats of their own, and moving into a new tool should not mean converting all of them first.

Glossaries now support native TBX. Lara Translate supports TBX (TermBase eXchange) for Glossary import and export through the SDK, so you can bring existing terminology databases into Lara Translate and export Lara Translate Glossaries in TBX format without first converting them to CSV. This improves interoperability with CAT tools and existing localization workflows. Support in the web app is planned for release in the next few days.

EPUB is now a supported Document Translation format. Document Translation supports EPUB files through the SDK, in addition to the Lara Translate web app. Developers can integrate EPUB translation directly into automated publishing and localization workflows without first converting files into an intermediate format, which makes EPUB a first-class format across both interactive and programmatic workflows.

Explore the Lara Translate developer documentation.

What to expect next

September’s shipped updates, the ChatGPT plugin, the Group Admin role, My Translations across web and mobile, Generative Image Translation in the web app, SSO and Meaning in Context on mobile, and TBX and EPUB support in the SDK, are built around one theme: making Lara Translate available in more places and easier to run at scale.

A few more things are already moving. The Lara Translate integrations for Claude, Slack, and Make are in marketplace review, TBX import and export is planned for the web app in the next few days, and work has started on a new Lara Translate web app design. You can follow the full changelog on the What’s New page.

FAQ

How do I use Lara Translate in ChatGPT?

Install the Lara Translate plugin from the ChatGPT Plugins Directory. Once installed, you can invoke it directly in a conversation, including through an @ mention or from the + menu where available. Availability can depend on your plan and workspace settings.

Are the Lara Translate integrations for Claude, Slack, and Make available yet?

Not yet. The ChatGPT plugin is published, and the Claude, Slack, and Make integrations are currently going through their respective marketplace review processes.

What is the Group Admin role, and how is it different from a Team Owner?

The Group Admin role, available for Team accounts, including Enterprise accounts that run on Team plans, lets a Team Owner delegate the day-to-day management of an individual Group to another person. A Team Owner owns and controls the entire Team account, while a Group Admin manages a single Group on the Owner’s behalf.

What is Generative Image Translation?

Generative Image Translation is a mode in the Lara Translate web app that recreates an image around the translated text, adapting typography, spacing, and layout for a natural result. You can choose it instead of Inpainting, which replaces the text while keeping most of the image unchanged, depending on the balance you want between cost, visual fidelity, and how much of the image is modified.

What is My Translations and where can I use it?

My Translations is a personal library where you can save text translations, including the source, the translated content, and your edits, then reopen, edit, or delete them later. It is available across the Lara Translate web app, iOS, and Android.

What is Meaning in Context?

Meaning in Context, available on iOS and Android, lets you view the meaning of an individual word directly within the text you are translating, with the explanation generated from the surrounding context. It is most useful for ambiguous words whose meaning depends on the sentence around them.

What is TBX support for Glossaries?

Lara Translate now supports TBX (TermBase eXchange) for Glossary import and export through the SDK, so you can move terminology databases in and out of Lara Translate without converting them to CSV first. Web app support is planned for the next few days.

Can Lara Translate handle EPUB files?

Yes. Document Translation supports EPUB files in the Lara Translate web app and through the SDK, so you can translate e-book and digital publishing content without converting it to an intermediate format first.

This article is about

• The Lara Translate plugin for ChatGPT, now published in the ChatGPT Plugins Directory
• Lara Translate integrations for Claude, Slack, and Make in marketplace review
• The new Group Admin role for Team and Enterprise accounts, and how it differs from a Team Owner
• My Translations, now live across the web app, iOS, and Android
• Generative Image Translation in the Lara Translate web app, alongside Inpainting mode
• Single Sign-On on iOS and Meaning in Context on iOS and Android
• Native TBX support for Glossaries and EPUB support through the SDK

L'articolo Lara Translate September 2026 Updates: Lara Translate in ChatGPT, Group Admins, and My Translations proviene da Blog | Lara Translate.

This is the everyday reality of agency and in-house translation teams, and it is exactly the gap the new Lara Translate Wordscope integration is built to close. By adding Lara Translate’s adaptive AI as a private machine translation engine inside Wordscope, the same context-aware suggestions reach every translator on the project while the shared translation memories, glossaries, and quality checks keep doing their job.

The problem: one project, eight languages, six translators, four days

Go back to Marta’s Tuesday. The manuals are technical, the client is particular, and the terminology is non-negotiable: a mistranslated safety warning or an inconsistent product name is the kind of error that gets an agency dropped. Her translators are skilled, but they are working in parallel, each pulling from the shared translation memory and glossary the project was set up with. The pressure is not on their language ability. It is on throughput and consistency at the same time.

Wordscope is where that work already happens. It is a cloud-based computer-assisted translation (CAT) platform, used by professional translators and teams across dozens of countries, that combines private translation memories, terminology databases (including access to official glossaries such as IATE, TERMIUM, and UNTERM), quality-control checks, and collaboration tools in the browser. Wordscope solves the coordination and review side of the job well. What agency project managers keep asking for is a way to give every translator a strong, context-aware first draft to edit, rather than starting each segment from scratch, without introducing a separate tool that ignores the memories and terminology the project depends on.

That is the specific gap. Marta does not need another standalone translation app. She needs the productivity of adaptive AI to show up inside the environment her team already works in, governed by the same linguistic assets. That is what the integration delivers.

What the Lara Translate Wordscope integration actually does

The integration lets Wordscope users add Lara Translate as a private machine translation engine. Once connected, Lara Translate’s engine produces intelligent, context-aware translation suggestions directly inside the Wordscope editor, alongside the existing workflow rather than in place of it.

Three things make that meaningful for a production team. First, the suggestions are context-aware: Lara Translate combines large language model reasoning with neural machine translation, so it weighs the surrounding sentence and domain instead of translating word by word. Second, the integration preserves Wordscope’s own assets: your private translation memories, terminology databases, and quality-control tools keep operating exactly as before, so the AI accelerates the draft while your controls guard the output. Third, the coverage is broad, spanning 200+ languages and working across all of your Wordscope projects and the file formats the platform already supports, from Word, Excel, and PowerPoint to HTML, InDesign, and SRT subtitle files.

In practice, a translator on Marta’s project opens a segment, sees a Lara Translate suggestion that already reflects the sentence’s context, checks it against the shared memory and glossary, and edits rather than composes. The project manager gets faster first passes without loosening the terminology grip that keeps the client happy.

How the integration helps agency and in-house teams

Consistency stays under your control

The suggestions coexist with Wordscope’s private translation memories and terminology databases. The AI proposes; your assets and QC checks dispose. For a team running one glossary across several freelancers, that is the difference between consistency by hope and consistency by system.

Delivery gets faster

Automated AI suggestions sit alongside the review workflow, so translators spend more time editing and less time drafting from zero. Across a large multi-language job split over a team, small per-segment savings compound into the kind of turnaround that lets an agency say yes to a tighter deadline.

The learning curve is close to zero

Translators keep working in Wordscope, the tool they already know. There is no new editor to learn and no export-import shuffle between applications. The engine simply becomes another suggestion source inside the interface, which is what makes it realistic to roll out across a whole team rather than one power user.

How to set up the Lara Translate Wordscope integration

Connecting the two takes two stages: create your API credentials in Lara Translate, then register Lara Translate as a private MT engine in Wordscope. You will need an active Lara Translate account with API access and your Wordscope settings open.

Step 1: Generate your Lara Translate API credentials

In your Lara Translate account, open My Account, then API, then Credentials. Click Create new credentials and give them a recognizable name (for example, “Wordscope”). Select Create, then immediately click Download Credentials and save the file somewhere secure. This step matters: your Access Key Secret is shown once, at creation, and cannot be retrieved afterward. If you lose it, you will need to create a new key. This gives you two values, an Access Key ID and an Access Key Secret, which Wordscope will ask for next.

Step 2: Add Lara Translate as a private MT engine in Wordscope

In Wordscope, open Settings and go to the Machine Translation tab. Click Add a private MT engine, select Lara Translate from the list, and enter your Access Key ID and Access Key Secret. Click Confirm. Once Wordscope validates the credentials, Lara Translate is live across all of your Wordscope projects, and its suggestions start appearing in the editor for your team.

That is the whole setup. For teams, one project manager typically creates the credentials and configures the engine, and every translator on the workspace works with the suggestions from there.

For a busy agency, the value of the Lara Translate Wordscope integration is that it adds speed to a workflow without adding risk to it. The AI raises the floor on every first draft, and Wordscope’s memories, glossaries, and checks keep the ceiling exactly where the client expects it. Marta’s Tuesday is still a big job, but it is now a job her team can deliver on time and on terminology, which is the whole point.

Frequently asked questions

What is the Lara Translate Wordscope integration?

It is a connection that adds Lara Translate as a private machine translation engine inside Wordscope. After you set it up, Lara Translate’s context-aware AI produces translation suggestions directly in the Wordscope editor, while Wordscope’s private translation memories, terminology databases, and quality-control tools continue to run as usual.

Do I need a paid Lara Translate plan to use it in Wordscope?

You need a Lara Translate account with API credentials. The free plan lets you create a single API key with a limited monthly character allowance, while the Pro and Team plans raise the rate limits and volume for production work. For an agency running high-volume, multi-language projects, a paid plan is the practical choice.

Where do I find my Lara Translate Access Key ID and Access Key Secret?

In your Lara Translate account, go to My Account, then API, then Credentials, and click Create new credentials. After you name and create the key, click Download Credentials right away and store the file securely. The Access Key Secret is displayed only once at creation and cannot be recovered later, so if you lose it you will need to generate a new key.

Does the integration replace my translation memories and glossaries in Wordscope?

No. The integration preserves Wordscope’s private translation memories, terminology databases, and quality-control tools. Lara Translate adds AI suggestions on top of that setup, so your existing linguistic assets keep governing consistency and your review process stays in place.

How many languages does the integration support?

Lara Translate supports 200+ languages, and the integration makes those available inside Wordscope across all of your projects. It also works with the file formats Wordscope already handles, including Word, Excel, PowerPoint, HTML, InDesign, and SRT subtitle files.

Is my content kept private?

Lara Translate is built by Translated, an Italian, GDPR-compliant company, and it offers a zero-retention Incognito mode for sensitive material as well as a data processing agreement for organizations that require one. Wordscope’s own private memories and workspace controls apply to your projects as well.

This article is about

• What the Lara Translate Wordscope integration is: Lara Translate added as a private machine translation engine inside the Wordscope CAT tool.
• Who benefits most: language service providers and agency or in-house teams that need faster delivery without losing terminology consistency.
• What the integration does: context-aware AI suggestions in the Wordscope editor across 200+ languages, alongside existing memories and glossaries.
• How it helps: faster first drafts, controlled consistency, and a near-zero learning curve because translators stay in Wordscope.
• How to set it up: create API credentials in Lara Translate, then add Lara Translate as a private MT engine in Wordscope’s Machine Translation settings.
• What to watch: the Access Key Secret is shown only once at creation, so download and store it immediately.

Sources

• Lara Translate + Wordscope integration
• Using Lara Translate in Wordscope (setup guide)
• How to generate an API key for Lara’s API
• Wordscope (CAT platform)

This article was produced by the Lara Translate content team. Lara Translate is an AI translation platform built by Translated, with more than 25 years of professional translation experience. Language service providers and translation teams use Lara Translate inside Wordscope and other CAT tools to deliver faster across 200+ languages and 60+ file formats, while keeping their translation memories, glossaries, and quality control firmly in charge.

L'articolo Lara Translate + Wordscope: The AI Machine Translation Integration for Faster Project Delivery proviene da Blog | Lara Translate.

That guess is the problem. General AI models translate without knowing your industry, your terminology, or the difference between two meanings of the same word. Lara Translate now connects directly to Langdock through the Model Context Protocol (MCP), so your teams get specialized, context-aware translation inside the agents and workflows they already use. No new tool to open. No API keys to manage.

The use case: one AI assistant, dozens of languages

Langdock is an enterprise AI platform used by more than 10,000 companies, including Merck, BASF, Würth, and Trumpf. At Merck alone it reports more than 33,000 monthly active users. When AI adoption reaches that scale, translation stops being a specialist task handled by a localization team and becomes something thousands of employees do casually, several times a day.

That is exactly where general translation falls down. An operations manager translating a compliance memo has no way of knowing the assistant rendered the same legal term three different ways across three paragraphs. A support lead answering a German ticket does not catch that the tone landed wrong for a formal customer. Each case looks fine on its own. Multiply it across an entire workforce and the inconsistency becomes a real cost: reworked documents, mixed terminology in the same report, and messages that quietly undermine how the company sounds in each market.

Until now, an ops team using Langdock had two options, and neither was good. The first was to paste the text into the general model inside Langdock, which guesses at context, uses different terminology each time, and remembers nothing about past choices. The second was to leave Langdock, open a separate translation tool, and paste the result back, which breaks any automation and, for regulated industries, raises the question of where sensitive content was sent. The integration removes that trade-off.

What the integration does

The integration connects Lara Translate to Langdock over the Model Context Protocol, an open standard that lets a platform like Langdock call an outside service directly. In practice, the Lara Translate MCP server acts as a bridge between Lara’s translation models and your Langdock environment. Once it is connected, Lara’s actions show up as tools that any agent or workflow can use: Translate, Detect language, List languages, List memories, Create memory, and more.

That means translation becomes a step you can build into automations, not a manual copy-and-paste job. A support workflow can read an incoming email, detect its language, translate the body with Lara, and draft a reply, all without leaving Langdock. An agent that summarizes weekly reports can pull in updates written in four languages and return one clean summary in yours.

The quality difference comes from what Lara is. It is built only for translation, trained on 25 million human-translated documents, and designed to read context rather than translate word by word. Ask a general model to translate “terra” in a text about tennis, and it may return “earth”; Lara understands the sport and returns “clay.” On top of that, Lara applies your translation memories and glossaries, so a term your company has already agreed on stays the same everywhere it appears. For teams with strict data requirements, Lara Translate operates under GDPR, and its Incognito Mode processes translations without keeping any history.

How it helps

For an enterprise ops team, the practical gains are direct. Translation lives where the work already happens, so people stop switching between tools. Terminology stays consistent at scale, because memories and glossaries carry the same choices across every agent and every user. Quality holds up on real business content, since Lara reads domain and context instead of guessing. The connection itself is simple and controlled: it uses OAuth 2.0, so there are no API keys to distribute or rotate, and you can turn on an “Ask for confirmation” step if you want a human to approve actions before they run. And it covers the full range of content a company actually moves: emails, documents, support tickets, reports, marketing copy, and even code comments.

How to set it up

You need two things before you start: an active Langdock account and an active Lara Translate account. From there, the whole setup takes a few minutes.

1. In Langdock, click your profile icon in the bottom-left corner and select Integrations.
2. Click Add integration, then Start from scratch, and choose Connect remote MCP as the type.
3. Enter the Lara Translate server URL: https://mcp-v2.laratranslate.com/v1
4. For authentication, choose OAuth 2.0 (Dynamic Client Registration), then click Create and connect.
5. When the login window appears, sign in to your Lara Translate account.
6. Select List Server Features to see the available actions, then enable the tools you want, such as Translate, Detect language, and List languages. You can optionally toggle Ask for confirmation for manual approval, then save.

Once you save, Lara’s translation actions are available to any Langdock agent or workflow. You can test it right away by asking an agent to translate a short phrase into another language.

The bottom line

When a whole company standardizes on one AI assistant, translation becomes everyone’s job whether they realize it or not. The Langdock integration means that job is handled by a system built for it: Lara reads context, keeps your terminology consistent, and works inside the agents and workflows your team already uses, without adding a tool or an API key to manage. For an ops team juggling supplier messages, tickets, and internal docs across markets, that is the difference between translation that quietly works and translation you have to check.

Frequently asked questions

What is the Lara Translate + Langdock integration?

It is a connection that brings Lara Translate’s translation engine into Langdock through the Model Context Protocol (MCP). Once set up, any Langdock agent or workflow can translate text across 200+ languages, detect languages, and use translation memories without leaving the platform.

Do I need an API key to connect Lara Translate to Langdock?

No. The integration uses OAuth 2.0, so you connect by signing in to your existing Lara Translate account when Langdock prompts you. There are no API keys to create, distribute, or rotate.

Which languages does Lara Translate support in Langdock?

Lara Translate supports more than 200 languages, covering the world’s most widely used languages for both text and document translation.

Where can I use Lara Translate inside Langdock?

Anywhere Langdock uses tools. After you enable the integration, Lara’s actions are available to any Langdock agent or workflow, so you can add translation as a step in an automation or call it directly from an agent.

How is this better than translating with a general AI model?

Lara is built only for translation and trained on 25 million human-translated documents, so it reads context and domain rather than translating word by word. It also applies your translation memories and glossaries, which keeps terminology consistent across every user and workflow, something a general model does not do on its own.

Is the connection secure?

The integration authenticates through OAuth 2.0 rather than shared keys, so there are no long-lived credentials to leak. Lara Translate operates under GDPR and offers an Incognito Mode that keeps no translation history. Langdock itself is built for enterprise use, with ISO 27001 certification, SOC 2 Type II auditing, GDPR alignment, and EU hosting for its application and most models.

This article is about

• Lara Translate now connects to Langdock through the Model Context Protocol, adding translation to agents and workflows.
• Enterprise teams on Langdock get context-aware translation across 200+ languages without switching tools.
• The connection uses OAuth 2.0 with no API keys, and setup takes a few minutes.
• Lara reads context and applies your memories and glossaries, so terminology stays consistent company-wide.
• Full setup instructions live in the Lara Translate help center, linked above.

Sources

• Lara Translate in Langdock via MCP (Lara Translate help center)
• Lara Translate + Langdock integration page
• Getting started with the Lara Translate MCP server
• Langdock security and compliance (ISO 27001, SOC 2 Type II, GDPR, EU hosting)
• Langdock (official site)

This article was produced by the Lara Translate content team. Lara Translate is an AI translation platform built by Translated, with more than 25 years of professional translation experience. Organizations running Langdock use Lara Translate to translate emails, documents, tickets, and reports across 200+ languages, directly inside their AI agents and workflows.

L'articolo Lara Translate + Langdock: Context-Aware Translation Inside Your AI Workflows proviene da Blog | Lara Translate.

August 2026 puts most of its attention there. Team owners can now organize members into groups and share glossaries and translation memories with a whole team or a single group. On mobile, iOS gained translation memories and glossaries, and Alternative Translations arrived on both iOS and Android. On the developer side, Audio-to-Text translation is available through the API and SDKs, and Audio-to-Audio translation added voice cloning. And Lara for Zapier is now live in the Zapier marketplace.

Here are the Lara Translate August 2026 updates: what shipped, and why it matters for teams.

What’s new for teams: groups, shared glossaries and translation memories

When a whole organization works from one Lara Translate account, the marketing team’s glossary is not the legal team’s glossary, and an owner needs a way to keep those apart without spinning up a separate account for every department.

Team owners can now manage groups directly inside Lara Translate. From the group management area, you can add a group, rename it, or delete it. Members can be assigned to groups individually or in bulk, and invitation links can be configured so a new user joins a specific group as soon as they accept. A member belongs to one group at a time.

Groups now connect to Lara Translate’s linguistic resources too. Glossaries and translation memories can be shared through a Shared with picker at three levels: personal (just you), the entire team, or one or more specific groups. That gives each part of an organization control over which terminology and translation history it works from.

At the same time, team members can create their own personal glossaries and translation memories and rename the resources they own, so personal work stays separate from shared team assets. The translation interface was updated to make personal, team, and group resources easier to tell apart when you select them.

See the full guide to team groups and resource sharing.

What’s new on mobile: translation memories, glossaries, and Alternative Translations

The translation you do on your phone should not forget everything your account already knows. Until now, memories and glossaries lived on the web.

The iOS app now brings translation memories and glossaries into the mobile experience. You choose which memories and glossaries Lara should adapt to from the options menu, and the selection applies across Text, Documents, and Images. It is account-wide, matching the web app, so the resources you pick follow you across surfaces. A visual indicator on the control shows when memories or glossaries are active, and resources that are still importing or that failed to import are disabled and show their status. Creating and importing resources still happens in the web app for now, with a Manage action that opens the right web page from iOS.

Alternative Translations is now available on both iOS and Android. Tap ALT below a translated text to see alternative renderings for specific words or phrases, each with a short explanation of the context or register, then swap in the one that fits. It is a way to fine-tune wording without rewriting the sentence or starting the translation over.

Learn how to use translation memories and glossaries in the mobile app.

Get Lara Translate on your phone

Translation memories, glossaries, and Alternative Translations, now in your pocket.

Download for Android | Download for iOS

What’s new for developers: Audio-to-Text and voice cloning

Spoken content is hard to work with. A recorded interview or a voice note is not searchable, not translatable in place, and not something you can drop into a text pipeline without transcribing it first.

Audio-to-Text translation is now available through the Lara API and SDKs. Upload an audio file and get back both the original transcription and its translation as text, in a single workflow. Supported formats are .wav, .mp3, .opus, .ogg, and .webm, with 68 source languages for the audio and over 200 target languages for the translated text. Files can run up to 200 MB and two hours, and processing is asynchronous, which suits longer recordings and backend integrations. You can apply translation memories, glossaries, and a translation style (Faithful, Fluid, or Creative). Pricing is €0.05 / $0.05 per minute of audio. It fits use cases like subtitles, translated transcripts, meeting notes, media analysis, and accessibility.

Audio-to-Audio translation now supports voice cloning. Instead of recreating translated audio with a predefined target voice, Lara can analyze the acoustic profile of the source speaker, including timbre, pitch, and vocal nuance, and generate target-language audio that closely mimics their original voice. You turn it on with the voice_cloning parameter. It requires explicit opt-in and has its own pricing, and it is built for audio with one primary speaker, such as voiceovers, narrations, training material, and podcasts. This keeps speaker identity consistent across languages, which matters for multilingual media, training content, and presentations.

See the audio translation documentation for developers.

Is Lara Translate available on Zapier?

Yes. Lara for Zapier is approved and now live in the Zapier marketplace, so you can add translation to your workflows without building a custom integration.

The integration provides four actions: Translate Text, Upload Document for Translation, Get Document Status, and Download Translated Document. You can connect Lara Translate to the apps you already use, so a new file in a shared drive or a new row in a spreadsheet can trigger a translation and route the result wherever it needs to go. Translation memories, glossaries, and styles carry into those Zaps, so automated translations follow the same terminology as the rest of your work. Even with incognito mode.

Explore Lara Translate on Zapier.

What else is new in integrations: Wordscope and Langdock

Lara Translate is most useful where you already work. Two new integrations bring it into a professional CAT tool and into an AI workspace, so you do not have to switch apps to get context-aware translation.

Wordscope is a browser-based CAT tool for professional translators. With the new integration, Lara Translate works as a machine translation engine inside Wordscope, delivering suggestions directly in your translation projects. If you have an active Lara Translate subscription, you add Lara as a private engine in Wordscope under Settings, then Machine Translation, using your API credentials (the Access Key ID and Access Key Secret from My Account, API, Credentials). Once validated, Lara is available across all your Wordscope projects. See the Wordscope integration.

Langdock is an AI workspace where teams build agents and workflows. Through Lara Translate’s MCP server, those agents can call Lara’s translation tools directly, including language detection, translation, and memory management. You connect it from Langdock’s Integrations settings by adding the remote MCP server URL, authorizing with your Lara Translate credentials over OAuth, and enabling the tools you want. From there, translation becomes an action your agents can run inside a workflow. See the Langdock integration.

What to expect next

August’s shipped updates, team groups with shared glossaries and translation memories, mobile memories and glossaries on iOS, Alternative Translations on both mobile apps, Audio-to-Text through the API and SDKs, voice cloning for Audio-to-Audio, and Lara for Zapier, are built around one theme: making Lara Translate work for teams and for the tools teams already use.

A few more things are already moving. My Translations is rolling out as a dedicated area where you can save useful source and translated pairs, including your edits, and return to them later without repeating the work. And the Lara Translate plugins for Make and Slack have been submitted for review in their respective marketplaces. You can follow the full changelog on the What’s New page.

FAQ

How do team groups work in Lara Translate?

Team owners manage groups from a dedicated group management area inside Lara Translate. You can add, rename, and delete groups, assign members individually or in bulk, and configure invitation links so a new user joins a specific group when they accept. A member belongs to one group at a time.

Can I share a glossary or translation memory with only part of my team?

Yes. Glossaries and translation memories can be shared through a Shared with picker at three levels: personal (just you), the entire team, or one or more specific groups. This lets different departments work from their own terminology and translation history.

Can I use translation memories and glossaries in the Lara Translate mobile app?

On iOS, yes. You select which memories and glossaries Lara should adapt to from the options menu, and the selection applies across Text, Documents, and Images. It is account-wide and matches the web app. Creating and importing resources still happens in the web app for now, reachable through a Manage action from iOS.

What are Alternative Translations?

Alternative Translations, available on iOS and Android, let you tap ALT below a translated text to see alternative renderings for specific words or phrases, each with a short explanation of context or register, and swap in the one you prefer without rewriting the sentence.

What is the difference between Audio-to-Text and Audio-to-Audio translation?

Audio-to-Text returns written output: it transcribes spoken content and translates it into text. Audio-to-Audio produces a synthesized voice file in the target language. Audio-to-Text supports .wav, .mp3, .opus, .ogg, and .webm files up to 200 MB and two hours, with 68 source languages and over 200 target languages, priced at €0.05 / $0.05 per minute.

What is voice cloning in Audio-to-Audio translation?

Voice cloning lets Audio-to-Audio translation preserve the original speaker’s voice instead of using a predefined target voice. Lara analyzes the source speaker’s acoustic profile, including timbre, pitch, and vocal nuance, to generate target-language audio that mimics their voice. It is enabled with the voice_cloning parameter, requires opt-in, has its own pricing, and is built for audio with one primary speaker.

What can I do with Lara for Zapier?

Lara for Zapier, now live in the Zapier marketplace, offers four actions: Translate Text, Upload Document for Translation, Get Document Status, and Download Translated Document. You can connect Lara Translate to the apps you already use to automate translation inside existing workflows, and translation memories, glossaries, and styles carry through.

Which new integrations does Lara Translate support?

Two new integrations shipped alongside Zapier. Wordscope, a browser-based CAT tool, can now use Lara Translate as a private machine translation engine through your API credentials. Langdock, an AI workspace, can connect to Lara Translate’s MCP server so its agents call translation, language detection, and memory tools directly inside workflows.

This article is about

• Team groups in Lara Translate, with members assigned individually or in bulk and invitation links that join a specific group
• Sharing glossaries and translation memories with a whole team or one or more specific groups
• Translation memories and glossaries in the Lara Translate iOS app across Text, Documents, and Images
• Alternative Translations on iOS and Android
• Audio-to-Text translation through the Lara API and SDKs
• Voice cloning in Audio-to-Audio translation
• Lara for Zapier, now live in the Zapier marketplace
• New integrations with Wordscope (a browser-based CAT tool) and Langdock (an AI workspace, via MCP)

L'articolo Lara Translate August 2026 Updates: Team Collaboration, Shared Resources, and Voice Cloning proviene da Blog | Lara Translate.