
Lara Translate
Lara Translate is Translated's professional AI translator, handling text, documents, images and audio across 200+ languages with the original layout preserved, plus glossaries, translation memories and a live voice interpreter. Its permanent free tier needs no credit card.
What is Lara Translate?
Lara Translate is the AI translation platform published by Translated s.r.l., the Rome company founded in 1999 whose structured data still records that founding year. Its pitch is unusually specific for the category: not a chatbot that happens to translate, but a model built for translation alone, claiming twice the accuracy of the largest generic LLMs and five times the speed. The claim rests on the training material. Lara was trained on millions of real professional translations rather than text scraped from the web, drawing on a corpus of 25 million translations by senior linguists that captures not only the final wording but the corrections, the errors and the reasoning used when two reviewers disagreed. That is why the product can explain the choices it made, offer several valid alternatives per segment, and ask a clarifying question when a gender or a context is genuinely ambiguous.
The lineage is displayed openly: Matecat in 2011, the first tool with adaptive machine translation; ModernMT in 2017, the first commercial application of the Transformer; ModernMT at 200 languages and cited by Gartner, CSA Research and IDC MarketScape by 2023; then Lara in 2024. ModernMT is now folded into Lara, with a migration window that closes at the end of 2026.
Five modes share one billing unit, the character. Text runs to 203 listed languages. Documents cover 60 to 70+ native formats and come back with the layout untouched. Images are read and rewritten with the design preserved. Audio files return as translated recordings. A voice interpreter handles live bilingual meetings in 186 languages. Three styles steer the register: Faithful for manuals, legal and technical copy, Fluid for general content, Creative for marketing and literary work. Glossaries, translation memories and learning from your edits keep terminology consistent over time.
It reaches you almost everywhere: web app, iOS and Android, a Chrome extension covering Gmail, Google Docs and Google Slides, a full API on every plan, a CLI with GitHub Actions, an MCP server for Claude Desktop, Cursor and ChatGPT, and native plugins for Trados Studio and memoQ. When the machine is not enough, a Translated linguist reviews the output. The vendor claims more than 350,000 customers across Lara and Translated.
What it does
- Translate text across 200+ languages, with automatic source-language detection
- Translate documents in 60 to 70+ native formats while keeping the original layout intact, from PDF, DOCX, PPTX and XLSX to InDesign, DITA, XLIFF, PO, subtitles and source code
- Translate the text embedded in images and get back either a finished image or structured text pairs
- Translate audio files and receive a translated recording, from MP3, WAV, OGG, WebM or Opus
- Interpret a live bilingual conversation in 186 languages by pressing and holding the microphone
- Enforce your own wording through glossaries, translation memories, added context and explicit instructions
- Escalate sensitive output to a professional Translated linguist for human review and certified translation
When to use Lara Translate / When not to
A quick filter to help you decide if Lara Translate is the right fit.
When to use Lara Translate
- Professional translators and MT post-editors, who plug Lara into Trados Studio, memoQ, Matecat or Wordfast and bring their existing translation memories and glossaries with them
- Localization managers running continuous pipelines through Crowdin, Locize or TranslationOS, with shared glossaries, shared memories and one pooled character quota for the whole team
- Developers and product teams that need translation as an API: text, document, image and audio endpoints, SDKs for seven languages, a CLI wired into GitHub Actions and an MCP server
- Marketing and content teams shipping multilingual campaigns, who need the Creative style, industry glossaries and in-place translation inside Gmail, Google Docs and Google Slides
- Legal, finance and document teams that occasionally need a human safety net: a Translated linguist can review the AI output for contracts, filings and official paperwork
When not to use Lara Translate
- Anyone shopping for a general-purpose AI assistant: Lara only translates and interprets, it does not draft, summarise or reason about your content
- Organisations that require self-hosting or on-premise deployment, since the service runs exclusively on the vendor's AWS infrastructure in the EU or the United States
- Buyers who want a money-back guarantee: fees are billed in advance and are non-refundable, with no credit for partial or unused months
- Teams planning to distil or train their own model on Lara's output, which the terms forbid outright
- Freelance translators looking for a marketplace to find clients: Lara is a production tool, not a place where work is brokered
How to use Lara Translate
A typical end-to-end flow, from setup to results.
- Open laratranslate.com/translate: text translation works immediately, with no account and no installation
- Pick the target language and either set the source language or leave automatic detection on
- Paste your text, or switch tab and upload a document, an image or an audio file
- Choose the register you need: Faithful, Fluid or Creative
- Add context, a glossary or a translation memory when specific wording has to be enforced
- Turn on Incognito mode before pasting anything confidential, so nothing is stored or used for training
- Read the result, compare the alternatives Lara proposes, and correct what you disagree with: in Learning Mode those edits feed back into the model
- Create a free account to unlock document translation, the higher monthly quota and the mobile apps
- For live meetings, open the interpreter, select the language pair, press and hold the microphone to speak, then release to hear and read the translation
- For technical use, generate an API key in your settings, then install the Trados Studio or memoQ plugin, the Chrome extension, the CLI or the MCP server
Pros & Cons
Pros
- A genuine specialist rather than a generalist that also translates: 25 years of trade behind it and a 25-million-segment professional corpus
- Unusually broad coverage for one subscription: text, laid-out documents, images, audio and live interpreting, all metered in the same characters
- Terminology actually holds, thanks to glossaries, translation memories, supplied context and learning from your edits
- Incognito mode keeps content out of storage and out of training, and it is not reserved for paying customers
- Compliance is above the sector average: published DPA, sub-processors named with their processing locations and legal bases, ISO 17100, ISO 27001:2022, ISO 9001:2015 and PCI badges
- Full API access on every plan, free tier included, with SDKs, CLI, MCP server and CAT-tool plugins rather than a paywalled developer track
- No friction to start: text translation works without an account, and the free plan asks for no credit card
Cons
- Nothing is refundable: billing is in advance, with no credit for partial months, downgrades or unused months, and refunds only at the vendor's discretion above 2% downtime
- The uptime figure written into the terms is 98%; the 99.9% target and a contractual SLA are Enterprise-only
- Published numbers disagree between pages: 61 versus 70+ file formats, 182 versus 186 interpreter languages, 200+ versus 203 languages
- Prices are shown on annual commitment by default, so the monthly-billing rate is not visible until you move the toggle
- One free account per person or entity, and the vendor reserves the right to change or withdraw the free plan without notice
- No self-hosted option: content transits the vendor's AWS estate in the EU or the United States depending on the case
- The group trust centre that all six certification badges point to refuses plain HTTP requests, which makes those claims awkward to verify from outside
Pricing & Plans
A permanent free plan is available at no cost, covering 60,000 characters per month plus 10,000 characters of API usage, and it requires no credit card. The cheapest paid tier is Pro at EUR 8 per month on annual billing, which raises the allowance to 500,000 characters per month. Annual commitment is advertised as up to 21% cheaper than monthly, and the terms state that fees are payable in euros, in advance, and are non-refundable.
- 60
- 000 characters/month
- one user
- documents up to 10 MB
- no glossary
- no extra usage
- free API up to 10
- 000 characters/month
- 500
- 000 characters/month
- one user
- documents up to 200 MB
- one glossary with unlimited entries
- extra usage EUR 0.25 per 10
- 000 characters
- API at EUR 20 per million characters
- 1.5M shared characters/month
- unlimited users
- shared glossaries and translation memories
- admin roles
- centralised billing
- premium support
- extra usage EUR 0.20 per 10
- 000 characters
- 1.5M characters EUR 25/month (best for 2-3 users)
- 3M EUR 45/month (4-6)
- 9M EUR 119/month (7-18)
- 20M EUR 219/month (19-40)
- 60M EUR 649/month (41-120)
- custom quota and overage pricing
- data sovereignty and residency
- enterprise compliance assistance
- Single Sign-On
- guaranteed SLA with 99.9% uptime
- dedicated account manager
- priority API
- payment by invoice
- free account with 10
- 000 characters/month
- then pay-as-you-go
- or annual volume tiers — Light EUR 600 (50M characters
- EUR 12/M
- overage EUR 15/M)
- Core EUR 2
- 000 (200M
- documents are billed per character with a 20
- 000-character minimum each
- images at 10
- 000 characters apiece
- audio at 40 characters per second
- the interpreter at 400 characters per second
Data, GDPR & hosting
A consolidated view of how Lara Translate handles your data.
GDPR overview
GDPR implementation is concrete and documented, not a badge. The privacy notice is issued under Article 13 of Regulation (EU) 2016/679 by a named controller, Translated s.r.l., Via Indonesia no. 23, Rome, VAT IT07173521001, info@translated.com; it is version 1.4, last updated May 2026. Legal bases are cited article by article: 6(1)(b) for contract performance, 6(1)(c) for accounting and tax duties, 6(1)(a) for cookies and marketing, 6(1)(f) for legitimate interest. Rights under Articles 15 to 23 are exercised at support@laratranslate.com, with a complaint route to the Italian supervisory authority. A Data Processing Addendum is published and folded into the terms, sub-processors are listed by name with their processing locations, and transfers rely on Standard Contractual Clauses and the EU-US Data Privacy Framework. Consent is managed through Cookiebot. No Article 27 representative is designated, and none is needed: the company is established inside the Union.
Who owns the data?
Users keep ownership of everything they submit, and stay responsible for holding the rights to it. For account data, the controller is Translated s.r.l. in Rome. For personal data sitting inside the content you send for translation, the roles invert: you are the independent controller and Lara acts as processor under its published Data Processing Addendum, with named sub-processors. What the vendor may do with your text depends on one switch. In Learning Mode you grant permission to use your content solely to refine the translation model, and nothing else; ownership stays yours and deletion can be requested from support. In Incognito Mode nothing is stored at all. Data is never sold on to third parties for their own marketing.
Reuse rights
You may reuse your source texts and your translations freely, without asking permission: nothing in the terms claims a licence over your output. Two limits are explicit and worth reading before you build on Lara. First, you may not use any output, data, feedback or metadata from the service to develop, train, fine-tune, validate or improve any machine-learning or large language model, competing or not, and you may not use it as training, evaluation or reference data for another product or model, commercially or otherwise. Second, if you use the share-by-URL feature, the source text, the language pair and any context you supplied travel inside the link, and anyone holding that link can read and edit them. Lara may also cite a customer by name and logo in its sales material, which the customer can refuse.
Data retention & training
Hosting summary
The controller is Translated s.r.l. in Rome, and disputes fall under Italian law before the Court of Rome. Hosting runs on AWS, which provides the application servers, translation processing, storage and backups, with the processing location declared as the EU or the USA under a data processing agreement with Standard Contractual Clauses and UK or Swiss addenda as needed. The sub-processor list, dated 3 December 2025, names every other party and where it operates: Stripe, PostHog, Cookiebot, ActiveCampaign and HubSpot in the EU; Sentry in the EU for EU data and in the US otherwise; Google Analytics 4, Google Ads and Google Workspace in EU data centres with extra-EU transfers covered by Standard Contractual Clauses and adequacy decisions; Meta and LinkedIn as joint controllers under Article 26. One caveat on geography: the domain resolves to a CloudFront edge node in Switzerland, which is a point of presence, not a place where anything is stored. Enterprise customers can negotiate data sovereignty and residency, and an ISO 27001:2022 badge is displayed site-wide.
Things to keep in mind
Risks and trade-offs to weigh before adopting Lara Translate.
- Over-trusting the output: the vendor itself reserves human review for legal, financial and official content, which is a quiet admission that the machine alone is not enough there
- Learning Mode stores what you paste in order to improve the model, so confidential text pasted without switching to Incognito mode has already been kept
- The share-by-URL feature embeds your source text and context in the link itself, and anyone who obtains that link can read and edit them
- Personal data inside the documents you upload stays your responsibility as controller: it is on you to check you were entitled to have it processed
- Skill erosion in teams that delegate everything, including the terminology check that used to catch the errors machines still make
- Despite the European base, processing can happen in the United States through AWS and several sub-processors, which matters for sensitive material
- Usage is metered in characters with automatic overage: one badly scoped batch job can burn a monthly quota, and a fluent translation is not automatically a faithful one — the Creative style deliberately moves away from the source
Setup & Integrations
Technical difficulty
For ordinary use there is nothing to set up: text translation runs in the browser with no account, and a free account without card details unlocks documents and the larger quota. The mobile apps and Chrome extension install in one step. Technical integration is more involved: generate an API key in settings, then pick an SDK from seven languages, the CLI with its GitHub Actions hook, or the MCP server. CAT-tool users install the plugin from the RWS AppStore or the memoQ store and paste the key. SSO and custom data residency go through sales.
Deployment
Apps stores
Integrations
Supported languages
Behind Lara Translate
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement Lara Translate.
Frequently asked questions
What exactly is Lara Translate?
What can Lara actually translate?
Is there a free plan, and does it need a card?
How much is the cheapest paid plan?
Will my content be used to train the model?
Is there an API, and on which plans?
Can a human check the translation?
Where is the data hosted?
Can I get a refund if I change my mind?
I use ModernMT — what happens to my account?
Should you pick Lara Translate?
Lara Translate is what a translation tool looks like when the company behind it has been doing translation for twenty-five years. It does one thing, and the narrowness is deliberate: this is not an assistant that also translates, it is a model trained on 25 million professional translations, complete with the corrections and the disagreements that produced them. That shows up in the details generalist models skip — explained choices, alternative renderings, a clarifying question when a gender is ambiguous, and terminology that actually holds because glossaries and translation memories are first-class rather than bolted on.
The breadth is the second argument. Text, laid-out documents, images, audio and live interpreting all draw on one character allowance, and the product reaches you through a web app, mobile apps, a Chrome extension, an API, a CLI, an MCP server and native CAT-tool plugins. Getting started costs nothing and asks for no card; the API is open on every tier, free one included, which is rarer than it should be. On compliance the vendor is more forthcoming than most: a published DPA, sub-processors named with their processing locations and legal bases, and an Incognito mode that keeps content out of training even for non-paying users.
The reservations are real but mostly commercial. Nothing is refundable, the uptime figure written into the terms is 98% rather than the 99.9% reserved for Enterprise, prices default to annual commitment so the monthly rate is hidden behind a toggle, and the marketing numbers contradict each other from page to page — 61 or 70+ formats, 182 or 186 interpreter languages. There is no self-hosted option either. For anyone whose work genuinely depends on translation quality rather than on getting the gist, Lara is a serious tool and the free tier makes the evaluation cheap.
- Choosing a selection results in a full page refresh.
- Opens in a new window.