- WAV: 44.1 kHz, stereo, 16-bit PCM.
- MP3: 320 kbps, encoded from the local audio.
- Mix uses your faders; an individual stem is a separate export.
Choose for the next step
Use WAV for another editing session, for keeping an uncompressed copy of the separated result, or when your next application requires PCM audio. Choose MP3 for a smaller listening or practice copy. Neither format makes a model estimate more faithful to the original performance.
What Stemix actually writes
The processing and export path uses 44,100 samples per second and two channels. WAV files contain 16-bit PCM samples. MP3 exports are encoded locally with a target bitrate of 320 kbps. A 48 kHz source therefore does not remain a 48 kHz export, and a 24-bit source does not remain 24-bit in the current WAV writer.
| Three-minute stereo file | Approximate audio data | Good for |
|---|---|---|
| 16-bit / 44.1 kHz WAV | 31.75 MB, plus a small header | Further editing and an uncompressed result |
| 320 kbps MP3 | 7.2 MB, plus encoding overhead | A smaller practice or listening copy |
These use decimal megabytes. They describe a single file, not all stems together. Silence, encoder metadata and exact duration can make the actual MP3 size differ slightly.
A larger file is not recovered detail
Converting a low-bitrate MP3 to WAV avoids another lossy encoding step, but it does not restore information the MP3 discarded. Likewise, exporting at 320 kbps does not reverse separation artifacts. Start from the best authorized source you have, and avoid repeatedly exporting and reimporting lossy versions.
Know which save action you chose
A stem export writes one separated track. Mix combines tracks using their current levels, including muted tracks at zero. If you want a quieter singer in a backing track, save Mix after adjusting the fader. Saving Vocals alone does not produce that backing mix.
For a batch, “Save all” creates one mix for each input file in a ZIP, using each file’s remembered levels. It is not a ZIP of every individual stem. Keep track of the selected file when making adjustments, and check one exported result before doing a large batch.
Leave room for peaks
Summing tracks can exceed the available sample range. The WAV writer clips samples outside that range rather than providing mastering or loudness normalization. If the export sounds distorted, lower the relevant track levels and export again. Use the Audio Inspector for sample-level checks, while remembering it is not a true-peak or LUFS meter.
This guide describes the current product controls and implementation. Separation results depend on the recording, device and model; examples are not quality guarantees.
Further reading: Spleeter · Web Audio API · How the tools work