Test RVC models on the DDLC character Sayori, via various hyperparams and datasets.
Note: was of a request. Since Sayori had no canon voice, these models were "interpretations" of her voice done via "experimental voice mixing": fine-tuning datasets were composed of various voices, on which versions were trained in various portions of.
ct-s1 - cts3 (11/07/2023)
- Trained on dataset of mixed voices, of around ~10 10 second clips
- Trained for ~100 epochs
ct-s4, cts4 (11/21/2023)
- Trained on dataset of mixed voices, of around ~10 10 second clips
- Trained for ~100 epochs
More info to be added.