Documentation
Subtitle workflow
Transcription, correction, translation, comparison, and subtitle export.
- New users
- Operators
A subtitle session operates on an audio or video source.
Normal stages
- Transcribe creates timed source-language cues.
- Correct creates a distinct revision for punctuation, wording, cue merges, and splits without translating.
- Translate creates a target-language revision and retains lineage to the source or corrected cues.
- Preview compares source, correction, and translation selections.
- Export writes selected cues as SRT or VTT, or concatenates a transcript.
Correction and translation are separate review decisions. Inspect the workflow snapshot and selected artifact IDs before continuing downstream. Selecting an older upstream revision can invalidate or clear dependent selections; use the reported impact rather than assuming downstream work is still current.
Speech recognition may run locally or through a provider depending on the configured engine. Translation may use an LLM or a translation provider. An exact plan should disclose remote data transfer and identify the selected input revision.