Split an audio file on split.tools and the pieces come out as WAV. That's a deliberate choice, and understanding it will save your audio from a slow, invisible death by re-compression.
The one-sentence version
WAV is a measurement; MP3 is a summary. WAV stores every sample of the waveform exactly; MP3 discards the parts a psychoacoustic model predicts you won't hear, in exchange for a file roughly a tenth the size.
What the numbers look like
Uncompressed 16-bit stereo WAV costs sample rate × 2 bytes × 2 channels:
| Format | Bitrate | One hour |
|---|---|---|
| WAV 16-bit / 44.1 kHz stereo | 1,411 kbps | ~635 MB |
| MP3 320 kbps | 320 kbps | ~144 MB |
| MP3 128 kbps | 128 kbps | ~58 MB |
That 10× size difference is real, and for distribution — podcast feeds, download links — MP3 (or AAC/Opus) is the correct choice. Nobody should stream your 635 MB WAV.
Why exports should still be WAV
The problem with lossy formats is generation loss. Each encode discards information; each re-encode discards information from an already-degraded signal. The damage compounds: decode a 128 kbps MP3, edit it, re-encode at 128 kbps, and you're two generations deep — audibly swishy in cymbals, speech sibilance, and reverb tails.
Splitting sits in the middle of a workflow: you'll typically edit, reassemble, or re-encode the pieces afterward. If the splitter exported MP3, it would silently insert an extra lossy generation into every workflow. Exporting WAV inserts zero — the samples decoded from your source are written out untouched. Cutting itself costs nothing: it selects samples, it doesn't re-encode them.
The practical rule:
Stay lossless through every intermediate step. Go lossy exactly once, at the end.
"But my source was already MP3"
Then the loss from that original encode is baked in — nothing recovers it. The goal now is not to add loss. WAV slices of an MP3 source are exactly as good as the source; MP3 slices would be strictly worse. And no, converting MP3 → WAV doesn't "improve" the audio; it just stops the bleeding.
Choosing the final format
When you do encode for delivery: spoken word is transparent to most listeners at MP3 96–128 kbps mono / 128–160 kbps stereo; music wants 256–320 kbps MP3, or AAC/Opus which achieve the same quality at lower bitrates. Every platform (Spotify, YouTube, podcast hosts) re-encodes uploads anyway — one more reason to hand them the highest-quality file you have.
Cut your audio losslessly in the browser-based splitter, keep the WAVs as masters, and compress once on the way out the door.