Repository navigation
Alex Floarea -ETERNAL - v104
Release containing v104 of Alex Floarea -ETERNAL- for DiffSinger
Changes from initial release:
- Trained with Reflow and Shallow Diffusion instead of DDSP for improved performance and quality.
- Trained with DiffSinger "Multi-Dict" branch for improved Cross-Language Synthesis.
- "Awaken" and "Electric" Vocal Modes trained with new datasets for improved stability and consistency.
- "Risen" Vocal Mode merged into "Dirt" due to similarity in tone and performance.
- "Allen" and "Bloodlust" vocal modes added.
- Support for Chinese, Korean, Thai, Spanish and French added.
Download the ZIP file and extract the folder within it to your "Singers" folder in OpenUTAU's directory.