Replace voice trainer source file

What is this?

With Replace source file you swap the file the voice trainer extracts the individual voice tracks from. This is useful when:

• You have received a better recording (such as an a cappella version instead of a noisy live recording)
• You have made corrections to the score and the MusicXML is now cleaner
• The original tracks were poorly recognised, for example because voices were too close together
• You want to switch to a different source format (from audio to MusicXML or vice versa)

How to replace the source file

1. Open the piece and switch to the voice trainer.
2. Click Replace file in the settings area.
3. Upload the new file — audio (MP3, M4A, WAV) or MusicXML/MXL.
4. Confirm the swap. Chorilo processes the file in the background.
5. Wait for the notification as soon as the new voice tracks are ready.

The old source file is permanently replaced. If you want to keep it, upload it beforehand under Add audio as a listening example.

What stays, what changes

Preserved:

• Loop points and jump markers
• Personal volume mixes of all members
• Voice line-up (soprano, alto, tenor, bass)
• Timbre of the voice tracks
• Tempo and transposition settings

Regenerated:

• The voice tracks themselves — that is, audio content for each voice
• The waveform display in the player

Permission

To replace the file you need the sheet_music.edit permission. Because the action triggers longer processing and affects all members, it is restricted to editors.

Which source file works best

• MusicXML from notation programs (MuseScore, Sibelius, Finale) — often the best results because voices are clearly separated
• A cappella studio recordings — very good when no accompaniment interferes
• Clean MIDI renderings — good, can sound a bit mechanical
• Live recordings with clean acoustics — possible, but reverb and audience make separation harder

Avoid recordings with loud instruments, lots of reverb or audience noise — the AI struggles with those.

Tips

• Before swapping, upload the old file as a listening example in case members still need it for comparison.
• Briefly inform members about the change — new sound quality is noticed immediately.
• Test the new file with a simple passage before pointing members to it.
• When the AI struggles, switching from audio to MusicXML often pays off — clean notation data delivers better results almost every time.