ijik-loker/

RVC_Gumball_Watterson_Japanese

Model Details

Model Description

• Developed by: ijik-loker
• Model type: Retrieval-based Voice Conversion (RVC)
• Language(s): Japanese

Uses

Training Details

Training Data

Voice clips dataset total duration

1. The Watch
2. The Void
3. The Vegging
4. The Test
5. The Tape
6. The Stories
7. The Slide
8. The Sidekick
9. The Safety
10. The Puppets
11. The Procrastinators
12. The Pest
13. The Nobody
14. The Misunderstandings
15. The Matchmaker
16. The Burden
17. The Best

Training Procedure

1. Remove noise using Ultimate Vocal Remover 5 UVR-DeNoise.
2. Extract vocals using RVC Web UI HP5-主旋律人声vocals+其他instrumentals.pth.
3. Remove echo and reverb using Ultimate Vocal Remover 5 UVR-DeEcho-DeReverb.
4. Manually diarise voices in Audacity using labels.
5. Export multiple to .wav by labels.
6. Train using RVC

• Target Sample Rate: 48k
• Version: v2
• Total training epochs: 200
• Base model G: f0G48k.pth
• Base model D: f0D48k.pth