File Transcription
The File transcription page lets you import audio files, run them through the active transcription engine, queue many files at once, and have folders transcribed automatically. It can also export subtitles when the engine returns segment timing.
Bring an audio file into TypeWhisper by dragging and dropping it onto the File transcription page, then transcribe it with the active engine. By default the same transcription model selected on the Dictation page is used, but you can pick a different engine and model just for file transcription with the per-file engine/model overrides.
Every file is converted to mono 16 kHz WAV via ffmpeg before transcription (video files have their audio extracted), so ffmpeg is required for all imports — if it is not installed, importing is disabled entirely and the page reports Unavailable: ffmpeg is not installed on this system. Install ffmpeg through your distribution to enable file transcription.
Supported input formats: wav, mp3, m4a, aac, ogg, flac, mp4, mkv, avi, mov, and webm. Video files (mp4, mkv, avi, mov, webm) have their audio track extracted.
You can queue multiple files for transcription at once. The batch queue tracks per-file status so you can see which files are pending, in progress, completed, or errored as the queue works through them. The engine/model override set for file transcription applies to every file in the queue, so you can run the queue on a different engine than your dictation model.
Watch folders transcribe new audio automatically as files arrive. Each watch folder supports the following options.
| Option | Description |
|---|---|
| Export format | Output format for the transcription: md, txt, srt, or vtt. |
| Output folder | Optional separate folder for the exported transcripts. Defaults to "Same as watch folder". |
| Engine / model override | Optionally transcribe this folder with a different engine and model than dictation. |
| Language override | Optionally force a language for files in this folder instead of auto-detection. |
| Auto-start on launch | Begin watching this folder automatically when the app starts. Starting is deferred and guarded, so a stale or unreachable watch path can't hold up app startup; if it can't start, the page reports why. |
| Delete source after export | Optionally remove the original audio file once its transcription has been exported. The source is deleted only after the export has committed, so a failed write can't lose the original. |
Exported transcripts are written to collision-safe paths and committed atomically, so re-processing a file or two files completing at once can't clobber an existing transcript.
Each watch folder keeps a processing history of roughly the last 100 entries, recording per-file success or error (with the error message on failure). Use Clear to empty the history.
When the active engine returns segment timing, you can export subtitles to SRT and WebVTT directly from the File transcription page. If the engine does not provide segment timing, subtitle export is not available for that transcription. See Transcription engines for which engines provide timing.
- Recorder — record longer WAV sessions and transcribe them after recording stops.
- Transcription engines — the engines that power file transcription.
-
HTTP API — transcribe files programmatically, using the same
ffmpeg-based importer.
| Date | Change |
|---|---|
| 2026-06-17 | Initial version. |
| 2026-06-17 | Added import and watch-folder screenshots. |
| 2026-06-17 | Removed unused screenshot placeholders. |
| 2026-06-17 | Corrected the ffmpeg requirement (required for all imports); added supported formats, per-file/per-watch engine-model overrides, watch-folder output folder, and processing history. |
| 2026-07-27 | v0.13.0: noted deferred/guarded watch-folder auto-start, atomic collision-safe transcript export, and delete-after-commit ordering for source files. |