The case for uncompressed
Most people who convert MP4 to WAV are not planning to listen to the result. They are feeding it to something: Whisper or another speech model, a forced aligner for subtitles, a DAW for a rough edit, a script that measures loudness. All of those work on samples, and all of them will decode a compressed file before they start.
Handing them an MP3 therefore adds a lossy encode in the middle of your chain and buys nothing. Better to extract WAV from MP4 once, work on that, and compress at the end if a compressed deliverable is what you need.
What it is, and what it is not
The audio track in an MP4 is virtually always AAC, which is lossy. Decoding it to WAV preserves exactly what that AAC decodes to, which means every artifact the original encode introduced comes along. Nothing is restored. There is no version of this operation that recovers the studio recording, and any tool suggesting otherwise is selling something.
What you do get is a stopping point: from here on, nothing further is lost. Cuts, fades, normalisation and analysis all operate on plain integers, and you decide when and whether a lossy encode happens.
Size, and the ceiling
Uncompressed audio is expensive. Stereo at 44.1 kHz and 16 bits is about 10 MB per minute, so the numbers get uncomfortable quickly. A five minute clip is 50 MB. A half-hour interview is 300 MB.
Input files up to 500 MB are accepted. Above that a browser tab cannot hold the container, and being told so before you start beats having the tab killed. For long material, trim the section you need first.
Everything happens on your machine. The engine is ffmpeg compiled to WebAssembly, downloaded once and cached by your browser afterwards.