MP4 to MP3
ffmpeg runs in this tab. Your video never leaves the device.
Up to 500 MB. The video is decoded here and only its audio is written out.
Checking whether the video and audio converter is already on this device
0%The short answer
The audio track is copied out and re-encoded as an MP3 at 192 kbit/s, which is where most listeners stop hearing a difference. The video is decoded in this tab by ffmpeg compiled to WebAssembly — a 31 MB download the first time you use it, and nothing on any visit after that.
The audio is pulled out and written as an MP3 at 192 kbit/s. The converter is a 32 MB download the first time and nothing after that, because it is kept on your device.
The formula
ffmpeg -i video -vn -c:a libmp3lame -b:a 192k out.mp3
Worked examples
- A talk or a podcast
- one minute is about 1.4 MB
- A music video
- audio only, video discarded
- A file already MP3 inside
- still re-encoded, at 192k
Reference table
| Bitrate | A minute is | Good enough for |
|---|---|---|
| 320 kbit/s | 2.4 MB | Archiving music |
| 192 kbit/s | 1.4 MB | Almost everything |
| 128 kbit/s | 0.9 MB | Speech, podcasts |
| 64 kbit/s | 0.5 MB | Voice notes, and it shows |
How to use it
- 01Open the page and wait a momentThe converter starts downloading in the background as soon as you arrive, unless your connection looks metered — in which case it waits until you pick a file.
- 02Drop the video inThe progress bar switches from downloading to converting. Long videos take a while: this is your own processor doing the work.
- 03Save the MP3The panel shows the size and how long it took, and on how many threads.
Also searched for
Searches that land here include mp4 to mp3, extract audio from video, video to mp3, convert mp4 to mp3 and rip audio from video.
Converting without uploading
What is actually happening
ffmpeg is the program almost every video service uses on the server. Compiled to WebAssembly it runs in a browser tab instead, at somewhere between a half and a quarter of native speed. That is slower than a server would be, and it is the whole point: the file stays on your machine, there is no queue, no size limit set by someone's hosting bill, and no copy of your video left in a temporary bucket.
Why the download is separate from the conversion
The progress bar does two different jobs, and it says which one it is doing. First it fetches the converter, which happens once per device. Then it converts, which happens once per file. Most tools of this kind show a single spinner for both, which is why people assume a 32 MB download is a stuck conversion.
Threads
Where the browser allows it, the threaded build runs on as many workers as your machine has cores, and the status line reports the number it actually got. That needs the page to be cross-origin isolated, which this route is; where a browser will not allow it, the single-threaded build is used instead and everything still works, just slower.
Questions people ask
Why is there a 32 MB download?
Because the alternative is uploading your video. That 32 MB is ffmpeg itself, compiled to WebAssembly so it can run inside the page. Every site that converts video without it is sending your file to a server, converting it there, and sending it back. This downloads the converter once, keeps it, and never sees your file at all.
Will it download again next time?
No. It is stored in the browser's Cache API, which is not the ordinary page cache and is not cleared when you close the tab. The second visit says “already on this device” and starts instantly. Clearing site data removes it, and then it comes back once.
How long does a conversion take?
Pulling audio out is fast — usually a small fraction of the video's length, because nothing has to be re-encoded except the audio. It depends on your machine rather than on a queue, so a laptop plugged in will beat a phone on battery, and the page tells you how many threads it managed to use.