Getting started

Install on Windows

1. Get the Windows package from Ilaria's Hugging Face repository. Check the instructions for your release before updating your NVIDIA driver or changing dependencies.
2. Extract the complete package to a writable folder. Run the launcher from that folder, so it can find the bundled runtime and audio tools.
3. Open ILARIA_RVC.bat. If startup fails, close that attempt and use ILARIA_RVC_LOGS.bat to see the console.
4. If the interface does not open automatically, use http://127.0.0.1:6969 on the same PC. Keep the launcher running while you work.

Convert a first clip

1. Import a compatible model from Other. Keep its optional matching index with the checkpoint.
2. Open Single Inference, select your model and add a short, dry voice recording.
3. Start with the default pitch extractor. Set pitch shift and Index Ratio to zero for a baseline.
4. Convert the clip and listen to the start, the loudest phrase and any consonants that sound softer than in the original.
5. Export a WAV reference. Follow Voice conversion to compare alternate settings on the same source.

Update and back up

For an existing installation, use Other → Update Ilaria RVC and read the release notes before applying the update. Save a copy of your datasets, exported models, training checkpoints and converted audio before switching versions. Keep training checkpoints separate from the inference models you use or share.

Use recordings and voice models you have permission to use. For model demos and availability, see the catalog.

Ilaria RVC source and installation notes · Release notes