uberdarwin/rvc-soul-coughing-trainer

🎤 RVC Soul Coughing Voice Trainer

An automated training pipeline for Soul Coughing vocals using RVC (Retrieval-based Voice Conversion).

Features

• —Automated Pipeline: Complete end-to-end training process
• —Soul Coughing Dataset: Pre-configured with Soul Coughing acapella files
• —Real-time Monitoring: Track training progress and logs
• —Easy Interface: Simple Gradio web interface

How to Use

1. 1.Upload Audio Files: Use the "Audio Files" tab to upload your training data
2. 2.Start Training: Click "🚀 Start Full Training Pipeline" in the Training tab
3. 3.Monitor Progress: Check the Status tab for training logs and progress
4. 4.Download Models: Once training completes, download your trained voice model

Training Process

The trainer automatically handles:

1. 1.Audio Preparation: Processes and cleans audio files
2. 2.Feature Extraction: Extracts F0, NSF, and ContentVec features
3. 3.Model Training: Trains the RVC model with optimal settings
4. 4.Progress Monitoring: Provides real-time training updates

Technical Details

• —Sample Rate: 40kHz
• —F0 Method: Harvest + DIO for pitch extraction
• —Feature Extraction: ContentVec for voice features
• —Training: Uses pretrained base models for faster convergence

Requirements

• —Audio files in WAV format (preferably isolated vocals)
• —Minimum 10 minutes of clean vocal audio
• —Consistent voice quality across all files

Notes

This Space is configured for Soul Coughing vocal training but can be adapted for other voice models by uploading different audio files.