Things you might want to ask

An AI translation workspace built around how documents actually feel. You bring in a file (or paste raw text), and it splits into blocks you can translate and edit one at a time — rather than handing back a wall of output. Glossaries, styleguides, sentence-level regeneration, and clean exports are part of the deal.

No. The Free tier gives you 1,500 credits a month — about five pages — with no credit card. You can translate end-to-end in Standard mode and use the full editor. Free also includes 5 lifetime messages with Literess; everything else (Premium mode, teams, batch, unlimited Literess) waits on a paid plan.

DOCX, PDF, MD, or TXT files via upload, pages from Notion, files from Google Drive, or just pasted raw text. The parsing into blocks is handled for you.

About 50, focused on languages where AI translation is consistently solid — including regional variants like Brazilian and European Portuguese, Latin-American and Castilian Spanish, Simplified and Traditional Chinese, and Cyrillic and Latin Serbian. The full list is in the language picker once you’re inside.

DOCX, Markdown, Google Docs, or back into Notion. Original formatting is preserved as best as possible — headings stay headings, lists stay lists, links stay links.

Standard is fast and accurate — 1 credit per source word — and is what most projects need. Premium uses a deeper model for nuanced material like literary fiction, poetry, or marketing copy with specific tone, and costs 3 credits per source word. You can switch modes per block, so you don’t have to commit one for the whole document.

Currently Google’s Gemini 3 Flash for Standard and Gemini 3.1 Pro for Premium. The choice is built into the mode, so you don’t have to think about model selection — but the model identity is visible per block, so you always know what produced a given translation.

Yes — that’s the point of the editor. Click any sentence, edit it inline, or regenerate just that one with optional directions like “make this more formal” or “match the rhythm of the previous line.” Everything else stays put.

Yes. Each block can hold several translation variants — useful when you want to compare a literal version with a freer one, or hand a client two options. One variant is active at any time, and that one is what shows up in exports.

No hard limit on document length. Long documents are split into blocks during ingest, and each block becomes its own translation job — so a 200-page book and a one-page memo go through the same pipeline, just at different scales.

Your client- or project-specific term list — character names that should never be translated, brand terms with a fixed rendering, technical vocabulary with chosen target equivalents. The model gets these terms with every translation request, so they come out consistently across the whole document.

Tone-and-voice rules in plain language: “Always casual, never formal.” “Translate gendered pronouns as ‘they’ when ambiguous.” “Use Oxford commas.” Whatever your style needs. The model uses it the same way it uses a glossary — every block, every sentence.

Yes. Glossaries and styleguides live at the workspace level. Attach them to as many documents as you want, and edit in one place — changes propagate to anywhere they’re used.

Yes. Per-document overrides let you toggle individual terms on or off, or swap a styleguide just for that project, without touching the underlying resource.

Not via CSV upload yet. But you can paste a list of terms in directly, or have Transept extract candidate terms from a document you’ve already worked with — then accept, reject, or edit them in bulk before they go live.

Three plans. Free (€0/mo, 1,500 credits) for trying it out. Starter (€19/mo, 100,000 credits) for regular work — unlocks teams, batch, premium mode, and unlimited Literess (who remembers your projects and preferences across documents). Pro (€49/mo, 300,000 credits) for higher volume with priority support. Top up any plan with credit packs that don’t expire. Full details on the pricing page.

A unit of translation budget tied to source words. Standard mode costs 1 credit per source word; Premium costs 3. So a 2,000-word document is 2,000 credits in Standard or 6,000 in Premium. Credits are only spent on AI translation actions — translate, regenerate, improve, or fix a block. Manual editing, exporting, leaving comments, and reading the document are free.

Existing translations stay where they are. New translation jobs pause until you top up. Buy a credit pack — Small (50K / €5), Medium (200K / €18), Large (500K / €40) — and pick up exactly where you stopped. Pack credits never expire.

No. Plan credits reset every billing cycle. Top-up credits from packs are different — those stick around until you use them.

Yes. Cancel from your account settings; you keep paid features until the end of the current billing period, then drop back to Free. No exit fees, no contracts.

Yes — a 7-day money-back guarantee on your first subscription payment, no questions asked. Email [email protected] within 7 days of your first charge and we’ll refund it to the original payment method (5–10 business days). Purchased credit packs are non-refundable after use, but pack credits never expire — so unused packs just stay on your account.

Yes, on Starter and Pro. Create a team, invite by email, and share documents, glossaries, and styleguides at the team level — rather than passing them around manually.

Yes. Real-time collaboration with block-level locks: when someone is actively editing a block, others see it as taken so you don’t overwrite each other. Cursors, comments, and resolved threads all update live.

On any block. Threaded conversations, resolve when done. You can also @-mention Literess in a comment to get a second opinion on a phrasing — she shows up in the thread and replies inline.

Yes — generate a public share link for the document. Anyone with the link sees a read-only snapshot of the translation; the link expires after 30 days. For deeper review (commenting, editing), invite them as a teammate instead.

Yes. Connect once via OAuth, pick a page, and we import it as a Transept document. Export the finished translation back into Notion as a new page when you’re done.

Same flow. Authorize Drive, pick a Doc or supported file, import. Export back as a Google Doc when finished.

Mostly, yes. Headings, lists, bold/italic, and links are preserved. Complex layouts — multi-column, embedded objects, custom fonts — may simplify, since the editor prioritizes text fidelity over layout fidelity.

Email and Google OAuth are both supported. Email sign-in uses one-time codes, not passwords — we send a code to your inbox and you paste it back. No password to remember or reset.

Stored in our database while you have the account. Sent to AI providers (currently Google’s Gemini family) only for the translation request itself. Removed when you delete your account. Full details are in the Privacy Policy.

From account settings, or by emailing [email protected]. Deletion removes your documents, translations, glossaries, and styleguides — irreversible, so export anything you want to keep first.

Check spam first. If the code still isn’t there, write to [email protected] with the email you’re trying to sign in with — we can resend or troubleshoot.

Yes — tell her your client’s brand rules or what novel you’re translating, and she keeps it as a durable note. Up to 20 per account; review or clear them in settings.

No. Transept doesn’t run its own foundation models, and the commercial AI APIs we use (Anthropic, OpenAI, Google Vertex AI, Groq, OpenRouter) don’t train on your prompts or outputs under the API tiers we’re on. Your text is sent for the specific translation request, returned, and stored only in your own document. The full breakdown lives on Security & Data.

Application servers and the Postgres database run on Hetzner in Germany — full EU data residency for everything you create. Translation requests temporarily transit through AI providers (mostly US-based) for the duration of the request itself, under SCCs and the EU-US Data Privacy Framework. Backups stay in the EU and are encrypted at rest.

From Settings → Danger zone → Delete account. The action is irreversible: your documents, translations, glossaries, and styleguides go with it. Anonymised consent-audit-log rows are retained (we can’t demonstrate consent without them) but they no longer link to a personally-identifiable account. Export anything you want to keep first — DOCX, Markdown, Google Docs, and Notion exports are all in the editor toolbar.

For a copy of everything you’ve uploaded plus your account metadata, email [email protected] from the address on your account. We respond within 30 days — for most accounts within 24 hours. Document content is exportable self-serve from the editor; the email path covers the bits you can’t reach yourself (account record, consent log, billing history).

They’re off by default. To revoke after granting: footer → Manage cookies → toggle Analytics off. PostHog stops firing immediately; we don’t need to restart anything. Server-side telemetry continues to capture anonymous operational metrics (latency, errors) but never under your distinct ID once you’ve denied analytics.

We use double opt-in — you only receive marketing emails after confirming via a one-time link, and you can withdraw any time from Settings → Email preferences. Every marketing email also has an unsubscribe link in the footer. Transactional emails (sign-in codes, billing receipts) keep going through; those aren’t marketing.

No. We don’t sell personal data, and we don’t share it for cross-context behavioral advertising. The only third parties that touch your data are the sub-processors listed at /subprocessors, each operating under a DPA.

Not yet. We follow SOC 2-aligned operational practices (access control, encryption, change management, incident response) but we’ve not undergone a third-party audit. Transept is not a HIPAA-compliant platform — please don’t submit Protected Health Information. We are GDPR / UK GDPR / CCPA-compliant by design and offer the full set of data-subject rights.

Email [email protected] from your account email and we’ll send our standard DPA within one business day. The list of sub-processors it flow-downs is published at /subprocessors.

Start with [email protected] — Mariia handles privacy inquiries personally. EU/EEA residents also have the right to lodge a complaint with their local data protection supervisory authority (contact details for every EU DPA are on edpb.europa.eu).