Larger = more accurate, slower, more RAM.
Auto-detect, or pick explicitly.
Whisper can translate non-English audio to English.
0 = deterministic, 1 = more creative.
Higher = slower but more accurate.
Process multiple files sequentially.
Affects transcript export format.
Supported: --language, --beam-size, --temperature, --translate, --model. One per line or space-separated.
First transcription downloads the model (~39–244 MB depending on selection).
whisper webCLI is a privacy-preserving browser application that runs OpenAI's Whisper speech-recognition model entirely on your device via Transformers.js and ONNX. No audio is uploaded anywhere. It is the second tool in the webCLI family, alongside ffmpeg-webCLI.