Browse the help center

How translation works

Maintained by Vitalii VlasiukCo-founder

Every block is translated with awareness of the document around it — the chapter, the headings, the recurring terms — so the long-document drift that generic tools hit around page 50 doesn’t happen here.

On this page

What are Fast, Standard, and Pro mode?

Three modes. Fast and Standard both cost 1 word per source word; Pro costs 3. You pick a mode per document and can switch it per block. The model identity is visible on each block, so you always know what produced a given translation.

Translation modes, what each is for, and what each costs
ModeCostBest forAvailable on
Fast1 word per source wordQuickest response, minimal reasoning. The default for new documents.Every plan
Standard1 word per source wordA touch slower than Fast, better at nuance.Every plan
Pro3 words per source wordExtended reasoning — literary fiction, poetry, or marketing with a specific tone.Starter and Pro
Pick a mode per document or per block.

Why does the rest of the document change one sentence?

Because each block carries the document’s context, terminology stays consistent and the register doesn’t slide from formal to casual halfway through. Add a glossary or styleguide and that consistency becomes a guarantee rather than a hope — see Glossaries and Styleguides.

Common questions

Which AI model does Transept use to translate?

Primarily Google’s Gemini family, with each mode running on a different model — Fast on the quickest one, Pro on one with extended reasoning. The model identity is shown on every block, so you can always see what produced a given translation. None of the AI providers Transept uses train on your content; see Privacy & data.

If I hit Translate again, will it redo the blocks I have already edited?

No. Translate always skips blocks that already have a translation, so it will not overwrite your edits and will not bill you for them a second time — it only picks up untranslated blocks. To deliberately redo a block that is already translated, use Regenerate instead; see Regenerate & rewrite with a direction.

Why isn’t my translation memory being used when I translate?

Check the mode first. Translation memory is off on Fast — speed comes first there — and on for Standard and Pro. If past wording matters for a run, switch that block or the document to Standard or Pro. See Translation memory.

Explore the features

Still stuck? Ask Literess in the app, or write to [email protected].