Settings → General
General settings covers your default transcription preferences, how subtitle lines are wrapped, and the colour theme. Everything saves automatically as you change it.
Default transcription settings
- Default language — pre-selected on Transcribe and other screens that transcribe.
- Default model — pre-selects the speech model. It lists only models you have installed; install them on Speech Engines.
- Videos transcribed at once — how many files Transcribe runs together, 1 to 4 (default 1). The app shares your processor threads between them. It is disabled with a graphics-card build of Whisper, where files run one at a time.
The language and model are chosen per run on the Transcribe screen; these only set what is pre-selected.
Subtitle layout
These control how subtitles the app generates are wrapped, whenever it writes a subtitle it produced:
- Line length — Short (about 32 characters), Medium (about 42) or Long (about 52).
- Maximum lines per subtitle — a single line, or up to two.
- Split long subtitles — when a block has too much text for the limit, split it into several timed entries instead of allowing extra lines.
- Force precise timestamps — prevents subtitles appearing too early or lingering during silence. Recommended.
The layout does not re-flow a file you only edit in the Tools; those keep your line breaks.
Transcription quality
- Voice Activity Detection (VAD) — filters non-speech before transcription. It can improve accuracy on videos with long silences but may produce empty output on quiet speech. It needs the Silero VAD model, installed on Speech Engines.
- Reduce hallucinations — only matters when you give a vocabulary hint. Whisper reads the hint as context, which helps it spell names but lets its timing drift onto a fixed one-to-two-second grid and can skip whole passages. Turn it on to put timing first: the hint is then ignored. Without a hint, Whisper always runs this way.
- Post-process subtitles — cleans raw output after transcription: removes speaker markers, fixes punctuation, drops repeated lines and fixes timing overlaps. Turn it off to see the raw output.
Appearance
Dark mode switches between the light and dark theme; the first launch follows your system.
FAQ
Where do I pick the language and model for a particular video?
On the Transcribe screen, for each run. The ones here are only the starting values.
My VAD switch does nothing.
It requires the Silero VAD model, which you install separately on Speech Engines.
Why is "Videos transcribed at once" greyed out?
You are using a graphics-card build of Whisper, which runs one file at a time.