Start with your recording
Upload an audio or video file, record your microphone, or import a direct HTTPS media link. Linked media needs to allow browser access. For YouTube, use a saved file that you own or are authorized to process.
Choose your model
Base balances speed and accuracy. Tiny is a good starting point for slower devices; Small needs more memory and a longer download. Compatibility mode uses WebAssembly. GPU mode requires a browser and device with WebGPU support.
Your first transcription
The first run downloads the selected Whisper model from Hugging Face. Model files are cached by the browser when storage is available. Your local media is decoded and transcribed on your device; after the model is cached, transcription can continue without a network connection while this page remains loaded.
Review and export
Check names, numbers and important quotes against your recording. Edit cue text or timing, add missing cues and undo deletions. Export TXT or JSON for text workflows and SRT or VTT for timed captions. Word supports broad Unicode text; PDF export rejects characters the built-in font cannot render.
Local history and cloud history
Local history belongs to this browser and device. Clearing browser storage removes it. Pro cloud history belongs to your signed-in account and can be accessed across devices. Choosing a cloud upload sends that media for processing.
Supported formats
Audio and video decoding depends on your browser. MP3, WAV and WebM generally have wide support. If decoding fails, convert the recording to one of those formats and try again. Free mode accepts up to 200 MB and 20 minutes per file.