Skip to main content

npx remotion whisper transcribev4.0.546

Transcribe a local audio or video file into a Caption[] array.
Transcriptions runs locally in Node.js using WebGPU.

Install the required packages before you run this command.

If not already done, the selected Whisper model will be downloaded automatically.

Transcribe a file
npx remotion whisper transcribe public/interview.mp4

Flags​

--output​

Path to the output file.
The default is:

  • <media-name>-transcript.json next to the input file with --raw
  • <media-name>-captions.json next to the input file otherwise

Use - to write to stdout. Then progress and diagnostics go to stderr.

Pipe Whisper's transcript to jq
npx remotion whisper transcribe interview.mp4 --raw --output - | jq -r '.text'

--raw​

Write the Whisper transcription result directly instead of Caption[].

--overwrite​

Transcription overwrites existing output files by default.
Use --no-overwrite or --overwrite=false to prevent overwriting an existing file.

--model​

The Whisper model to use. Defaults to small.en for English or small when --language is set.

--language​

The spoken language name or code, such as german or de. Required for multilingual models.

--binaries-directory​

The directory where the platform-specific binaries and libraries that Remotion needs are located. Those include an ffmpeg and ffprobe binary, a Rust binary for various tasks, and various shared libraries. If the value is set to null, which is the default, then the path of a platform-specific package located at node_modules/@remotion/compositor-* is selected.
This option is useful in environments where Remotion is not officially supported to run like bundled serverless functions or Electron.

--log​

One of trace, verbose, info, warn, error.
Determines how much info is being logged to the console.

Default info.

--quiet​

Suppress progress updates. -q is an alias.