Domain:
Termbase:
XLSX · XLS · CSV · TBX · XML
Cap:
Terms are sorted longest-first before being passed to the LLM. This ensures
multi-word expressions are matched before their component words. If your
termbase exceeds the cap, the shortest terms are excluded.
Every term injected into the prompt costs tokens on every single chunk
translation. A large termbase increases API cost and pushes the system prompt
closer to the model's context limit. At very high term counts this can cause
the model to truncate or ignore parts of the prompt, including style guide
rules. Set the cap to the minimum that covers your document's terminology.
multi-word expressions are matched before their component words. If your
termbase exceeds the cap, the shortest terms are excluded.
Every term injected into the prompt costs tokens on every single chunk
translation. A large termbase increases API cost and pushes the system prompt
closer to the model's context limit. At very high term counts this can cause
the model to truncate or ignore parts of the prompt, including style guide
rules. Set the cap to the minimum that covers your document's terminology.
No termbase loaded
If you update this file, reload it here to apply the changes.
Style Guide:
PDF · DOCX · TXT · RTF
No style guide loaded — using basic translation rules
If you update this file, reload it here to apply the changes.
Input bilingual file
Drop your file here
SDLXLIFF · MQXLIFF · XLIFF 1.2 / 2.0
File Info
No file loaded yet.
Chunk:
Chunk size — how many segments the AI translates per API call
The AI writes all translations in a chunk into a single reply.
Long segments produce longer translations,
so large chunks risk hitting the reply size limit.
10 — Safest. Best for long or complex segments where each
translation is large. If a batch fails, little work is
lost. Also good for unstable connections.
20 — Recommended. A reliable middle ground for typical
mixed content.
50 — Fastest. Best for short segments where replies stay
compact. Risk: if a batch fails, more segments need
redoing at once.
The AI writes all translations in a chunk into a single reply.
Long segments produce longer translations,
so large chunks risk hitting the reply size limit.
10 — Safest. Best for long or complex segments where each
translation is large. If a batch fails, little work is
lost. Also good for unstable connections.
20 — Recommended. A reliable middle ground for typical
mixed content.
50 — Fastest. Best for short segments where replies stay
compact. Risk: if a batch fails, more segments need
redoing at once.
Lang:
→
(auto-detected from file, if available)
Filename:
Output:
Translation Output
Translations will appear here…