Control

The same names and voice from page 1 to page 500

Start translating

First 5 pages free. No card.

Literess, the AI editor in Transept

You: Make Grandma grumpier.

Literess: Done. She doesn’t just say it anymore. She grumbles it. sagte → raunzte

Ask in plain words. You approve every change. Meet Literess →

Interactive demo — click anything
Try
Before the first paragraph✦

Transept reads your document before it translates a word

Upload a document and Professional mode reads it first. It lists the terms and names it uses with how to carry them into your language, and writes down the voice it is written in. You check both, and every paragraph follows them.

Recorded in Transept
A Transept press release, set up for a German translation. Real recording, sped up.

What keeps a long translation consistent?

You set up the glossary, the memory and the style guide once for a book or a client. Every translation, regeneration and proofread reads them, in one language or twelve.

  1. A glossary for names and terms

    Add source and target pairs by hand, upload a CSV, or have Transept draft a glossary from a document you’ve already translated. Every term goes into every request, so a name on page 500 matches page 1.

  2. A memory of your decisions

    Every block you approve goes into memory automatically, with the alternatives you rejected and the comments that explained why. The next run gets the closest matches as reference, so a sentence you settled once stays settled.

  3. A style guide for the voice

    Write one yourself or have Transept generate it from a page of finished work: tone, register, and how to treat idioms and dialogue. When you regenerate a sentence with “more formal” or “keep the metaphor”, the style guide decides what that means for this text.

  4. Smart Proofread catches drift

    After translation, Smart Proofread re-reads the text against your glossary, your style guide, and the source, and fixes the blocks that drifted. Review changes lists every fix beside the earlier text, and you keep or revert each one.

  5. Literess reads like an editor

    Literess reads the translation the way an editor would and leaves inline comments where the tone has gone neutral, the meaning has drifted, or a term has slipped. She remembers your project between documents and acts on what you approve.

  6. One set per book or client

    A glossary or style guide can belong to one document, a project, or your whole account, so one client’s terms won’t leak into another client’s work. Teams share them.

In context✦

A generic AI translator treats your character’s name as a word it can paraphrase, and by page 50 the protagonist has three names. The same happens to a product term across forty pages, or to a legal definition that has to stay exact. The model translates each passage well on its own and forgets what it decided two chapters ago. Transept gives it that memory: the glossary pins names and terms, the translation memory carries the decisions you approved and the reasons behind them, and the styleguide holds the voice. All three go into every request, and Smart Proofread checks the result against them afterward.

Where is consistency the whole game?

Fiction with recurring characters

Aurelia stays Aurelia. The dialogue stays in voice. The detective’s catchphrase doesn’t shift in the third chapter.

Terminology drift

The product term that started as “tray” and became “panel” by page 40. Caught.

Register slip

The chapter that started formal and slowly turned casual. Caught.

Trust and control✦

Built for work you answer for

  1. Private by default

    Your text is never used to train AI, not by us and not by any provider we use, and it stays out of our logs. Delete your account and your text is deleted with it.

    How we handle data
  2. Security you can check

    Hosted in Germany, in Hetzner’s ISO 27001-certified data center. Every sub-processor is listed, and a DPA is one email away.

    All sub-processors
  3. Give it context

    Glossaries, style guides, translation memory and per-string notes go with every request. You set them up once per project.

    Glossaries
  4. You’re the human in the loop

    The AI drafts; you decide. Review every change side by side, keep or revert it, and send a review link when a client needs to look.

    Review links
  5. API, MCP and webhooks

    The full API on every paid plan. Connect Claude or another AI assistant over MCP, or start runs from your own tools.

    Developer guide
  6. Something bigger?

    Large catalogs, custom workflows, a DPA for procurement. Transept stays self-serve; write to us for the rest.

    support@transept.ai

✦From our labs

4× cheaper model, matched

Give a model the decisions behind past translations, not just their final wording, and it stops guessing. With enough of that context, a model costing a quarter as much matched a premium one on our own benchmarks.

The memory bet

FAQ

  • A glossary pins individual terms, such as a name, a product word, or a phrase. Translation memory matches whole segments: sentences and paragraphs you’ve already translated. The glossary keeps words consistent, and the memory saves you from translating the same sentence twice. Most professionals use both.
  • You can upload a CSV, paste from a spreadsheet, or build a draft automatically from a document you’ve already translated. The auto-build picks up recurring proper nouns and brand terms.
  • A style guide holds tone (formal, casual, literary), register, preferred sentence length, how to handle idioms, regional conventions, formatting rules, and things to avoid: anything you’d write down for a human translator.
  • If it misses a term, Smart Proofread is the safety net. It re-reads the document with your glossary at hand and fixes the blocks where a term slipped, and you keep or revert each change afterwards.
  • Yes: a small read fee on every word it checks (0.2 credits per word), plus your document’s mode rate on the blocks it fixes. Blocks that already read well get no fix and no fix charge, which is why the dialog shows the price as a range before it runs.
  • Translation memory is part of every plan, including Free, with no add-on to buy and no separate database to manage. Every translation you approve goes into your memory automatically.
  • Literess keeps up to 20 notes per account and also draws on your projects, preferences, glossary, style guide, and past translations. You can review or clear her notes whenever you like.
  • A glossary per client is the most common setup. Build one glossary per client (or per project) and switch on the ones each translation needs. Terminology never leaks between clients, and switching between projects takes one click.
  • They’re meant to work together. The glossary pins specific words, the styleguide shapes everything else, and both apply to every translation.
  • Smart Proofread checks glossary compliance (missed terms, inconsistent translations, accidental synonyms), styleguide compliance (register slips, tone drift, changes in sentence rhythm), terminology consistency across the document, the handling of numbers, dates, and units, and anything omitted or added compared with the source.
  • Trados and memoQ offer the best of classic CAT memory: a segment database with exact, fuzzy, and context (101%) matches. Transept differs in two ways. The first is matching: they stop at text similarity, while Transept adds semantic search, so it finds a sentence that means the same thing even when it’s worded differently. The second is where the memory lives: theirs is a panel you query beside the editor, while Transept’s builds itself from the work you approve and feeds the AI draft and the QA pass.
  • Literess is available on Free: your first 10 messages are free; after that, a message about your translation costs credits, as on every paid plan.
❧

Translate twenty pages and have them all read in your voice

No card · First 5 pages free