Plans, credits & the free tier

Three plans. Free is €0/mo with 5,500 credits and no card; Starter is €29/mo with 60,000 credits; Pro is €79/mo with 200,000 credits. Every plan includes the full editor, batch actions, workflows, and all three modes, Pro included. Paid plans add more credits, unlimited messages with Literess, unlimited teams, and API, MCP and webhooks.

• On the Free plan, you can translate a selection, page, whole document, or workflow against the same 5,500-credit balance. There is no separate batch cap. Translating starts on Pro and proofreading on Standard; translate on Standard too and the same credits go about twice as far.
• Creating a team and inviting people is free on every plan. Seats are optional: they fund a shared credit pool at €29/mo (Starter) or €79/mo (Pro) each, and every seat adds that plan’s credits to the pool. Without seats, each member’s work bills their own balance.

A credit is the unit of translation budget, tied to source words. Fast and Standard mode cost 1 credit per source word; Pro costs 3.

Chinese and Japanese text is counted per character: each Han or kana character counts as one word, the convention CAT tools use. Korean, which uses spaces, is counted by word.

Credits are spent when AI writes or reads your text: translating or rewriting a block, asking for more alternatives for a sentence, writing a fix, reading a document for a check, a glossary or a style guide, or laying out the pages of a PDF, a scan or a comic (PDFs, scans and comics). On Starter and Pro, Literess's answer to a message about your translation spends credits too. Editing by hand, the first alternatives for each sentence, leaving comments, exporting and translation memory are free.

Plan credits reset every billing cycle, and Settings → Billing shows the date of the next reset. Credit packs never expire, so anything you buy stays until you use it.

When a run costs more than your balance, the run dialog shows the smallest credit pack and the smallest plan that cover it, and offers Translate what fits (Run what fits for a run that doesn’t translate). It runs the opening part of the document that your credits cover, so even on Free you see your first pages translated. Your balance never goes negative, and the dialog tells you how many blocks that covered and how many more credits would finish the rest. On your first document the Translate dialog makes it simpler: next to the page you have open it offers What fits, which translates from the top as far as your credits go.

Existing translations stay put, and after you top up, the next run picks up at the first untranslated block.

Every paid action shows its price before you click: a small coin and a number on the button that runs it (Translate, Regenerate, More variants, Literess’s Confirm), the range on the note under Literess’s message field (point at it), or the cost line in the run dialog for a page, a document or a workflow. The number turns amber only when your balance can’t cover it (1 to 3 in the figure below).

The run dialog sets the price against your balance (“1,926 of your 5,500 credits”) and, for a run on Pro, gives the Standard price on the line under it. When a document opens for the first time, the price of translating all of it shows under Translate, with how far your balance reaches.

Your balance is always in view: in the editor’s top bar, and at the bottom of the sidebar on the dashboard, library and settings pages. Clicking it opens Settings → Billing.

Charges are whole credits, rounded down, and work that fails is never charged.

Settings → Billing shows where your credits went over the last 30 days, day by day, and lists every charge and top-up with the document it was for. A run is one line with its total, and Show each charge lists what each block cost.

Some runs can’t know their exact cost until they happen:

• A proofread bills a small read fee on everything it reads, but the model’s rate only on the blocks that actually need a fix.
• Translate always skips blocks that already have a translation.
• In large-parallel mode, the AI sync passes bill only when they update the shared glossary and style guide.
• A message to Literess bills the work her answer takes: a quick answer costs the floor, a long task more. A task that reaches the top of the range stops and asks before going on.

So the estimate shows a range instead of one number: the floor is the work that will definitely run, the ceiling assumes everything gets touched.

When you confirm, the ceiling is held from your balance so the run can always finish. As it works, you’re billed only for what was actually done, and when it completes, the unused part of the hold is returned automatically. You never pay the ceiling unless every block needed work.

If a run fails or is cancelled partway, only completed work is charged and the rest of the hold is released.