Transept

Most AI translation tools give you a good first pass and nothing else — the second document forgets everything the first one decided. Transept’s workspace runs the opposite direction: you bring in a document (DOCX, PDF, MD, TXT, or pasted text), translate block by block in a side-by-side editor, and every approved choice is written into a translation memory that carries forward. Glossaries built once toggle on per project and enforce terminology from the first page to the last across as many language pairs as the project requires.

The differentiating feature is how Transept stores decisions, not just outputs. When you approve a translation, the memory records the alternatives you rejected and the comments that explained the call. That means a translator returning to a project six months later — or a second translator joining mid-project — sees not just what was decided but why. For literary translation or brand-voice localization, where a single off-tone word can undercut a chapter or a campaign, that context is the work.

Transept fits professional translators managing client-specific styleguides, content teams batching multilingual publishing, and localization groups coordinating shared glossaries across reviewers. It does not fit teams that need to embed translation into a programmatic pipeline: the vendor provides no API and no self-hosted deployment path, so Transept lives entirely in the browser. Teams with automated localization workflows — strings piped from a repo, translated, and merged back without human staging — will need a different tool. The paid-only word top-up model also means high-volume batch jobs carry a cost that scales directly with source word count.