This transcriber was made to help students, researchers, and interviewers convert their audio files into text without needing a subscription, and without limits on audio file size or duration.
Disclaimer: Because this transcriber uses open-source models and APIs, its accuracy and speed are not guaranteed to be 100%.
Privacy and Performance
The transcriber runs on your device, using your browser's WebGPU or your device's own hardware. Since transcription happens on your end, your audio files are processed and stored only on your device and are never sent anywhere else.
Transcription may affect your device's performance while it runs, and its speed depends on your device's capability.
Modes
There are four modes in this transcriber:
- Fast (265 MB): The quickest mode for clear audio.
- Balance (604 MB): A balance of speed and recognition quality.
- High (1.52 GB): More recognition detail for challenging audio.
- Max (2.45 GB): The largest mode for the highest recognition quality; it uses the most memory and takes the longest.
Settings
In the settings, you can:
- Select an M-Vox mode.
- Choose the language you are transcribing. It is set to auto-detect by default, but for better accuracy you can set it to the language of your audio.
- Toggle auto-translate to English on or off. It is on by default.
- Toggle timestamps on or off for more accurate checking.
- Personalize the display. Under Personalization, you can pick from preset themes, change the font, and adjust the font size.
Tips
Since the transcriber is not 100% accurate, you can listen to the media player in the transcription display and follow the highlighted text. (Disclaimer: the highlight feature is sometimes out of sync with the media player.) You can then manually correct the transcript after downloading it.
