Free online audio tools for MP3 conversion, WAV conversion, extracting audio from video, audio merge, audio trim, audio compression, microphone recording, audio-to-text transcription, volume changes, playback speed changes, sample rate changes, bitrate adjustment, silence removal, and text to speech. Conversion, trimming, merging, compression, and audio editing run locally in your browser, while transcription can load an open-source Whisper model in the browser.
Audio and video are processed locally in your browser first; FFmpeg WASM and transcription model files are downloaded only when needed.
These tools run FFmpeg compiled to WebAssembly, so format conversion, trimming, merging, speed changes and resampling all happen in the browser instead of on a server. Audio-to-text runs an open-source Whisper model in-page, and stem splitting runs its separation model the same way.
The cost is that the compute is yours: transcribing or splitting a few minutes of audio can take a long wait, where the same file converts format in seconds.
Bitrate sets the balance between size and quality. Rough guidance for common MP3 settings; the real difference depends on your material and listening setup.
| Bitrate | Per minute | How it sounds | Suits |
|---|---|---|---|
| 96 kbps | about 0.7 MB | Intelligible, but detail is clearly gone | Speech only, meeting recordings, podcast drafts |
| 128 kbps | about 0.9 MB | Good enough for most purposes | Spoken content and background music |
| 192 kbps | about 1.4 MB | Close to transparent on music | General music distribution |
| 320 kbps | about 2.3 MB | The practical ceiling for MP3 | Music where you want maximum fidelity |
MP3 is lossy, and re-encoding cannot bring back high frequencies that are already gone. Converting a 128 kbps file to 320 kbps only makes it bigger; for fidelity you have to start from a lossless source.
FFmpeg in the browser reads the whole file into memory, so large inputs; especially extracting audio from a long video; can fail for lack of memory. Cut the source shorter when that happens.
Audio-to-text and stem splitting download a model and then do heavy computation. Time scales with audio length and device speed, and phones may not finish longer material at all.