- Who it helps
- Developers evaluating browser-based transcription
- Availability
- Public MIT code and demo. Model weights download on first use and have separate terms; supported paths differ by browser and model.
Plan for the first download
The README lists roughly 120 MB for the default Whisper Tiny and 206 MB for Base. It caches weights in browser storage for later offline use. Larger models can exceed browser memory.
Match the model to the browser
WebGPU uses the GPU through the browser; some paths fall back to WebAssembly. Threaded WebAssembly requires cross-origin isolation headers. Parakeet TDT requires WebGPU, so a fallback should not be assumed for every model.
What you can do
- Compare the documented inputs, access requirements and limits with your own task before trying it. Open reference ↗
Where the evidence stops
Source and example descriptions checked; no audio processed, network traffic audited or accuracy measured. No independent video demonstration verified.
Sources read. Public text checked November 14, 2026 UTC. No installation, model inference or full workflow executed.
Sources & further reading
GitHub · implementation
browser-whisper source ↗
MIT code; model downloads and licenses separate. Public page shows 234 stars. No transcription run.
Tanpreet Singh Jolly · page checked 2026-11-14
234 stars · observed 2026-11-14 13:09 UTC
Web · demonstration
browser-whisper examples ↗
File, cache and microphone example descriptions read. Playback and transcription not executed.
Tanpreet Singh Jolly · page checked 2026-11-14
Coverage note: Source and example descriptions checked; no audio processed, network traffic audited or accuracy measured. No independent video demonstration verified.
The record over time (1)
Dated source checks and published explanations. An observation date can differ from the original release date.
- · Added to Current
browser-whisper is a JavaScript library for turning audio or video into text on the user’s device.