When you actually need MP3, not a video
Three common cases drive most of the searches that land on this page. First: you recorded an interview or a meeting and you only want the audio for a podcast feed. Second: you are running a transcript through a service that wants audio (and the file is smaller without video, so it costs less). Third: you want to listen to a recorded talk on a phone with the screen off, which iOS will not do reliably from an .mp4. All three are easier with the audio extracted.
What quality you should expect
We export VBR MP3 at the LAME quality-2 preset, which lands around 190 kbps. That is the same range Spotify and Apple Music stream at, and indistinguishable from the source for speech, music, and screen-recording audio in any blind test. If your input audio was already low-quality (say, a Zoom recording), the output will not be magically better — but it will not be worse either.
Bigger files: use a desktop tool
Browser-based FFmpeg is capped around 500 MB for memory reasons. A typical hour-long video is well under that, but a 4K or multi-hour recording will hit the ceiling. For batch extraction or very large files, the Clipy desktop app handles it natively, or any standalone FFmpeg install does the job in one command.
Going to transcribe it?
If MP3 is just a pit stop on the way to a transcript, you do not need this tool — most transcription services accept MP4 directly. Whisper, Otter, AssemblyAI, and Rev all take video. The only reason to convert first is to cut upload time and cost, which can be worth it for hour-long files.