Multi Model RVC Inference

Multi-Model RVC Inference

Simplified RVC Inference for HuggingFace or Google Colab

Information

Please support the original RVC, without it, this inference wont be possible to make.

Features

• Support V1 & V2 Model ✅
• Youtube Audio Downloader ✅
• Demucs (Voice Splitter) [Internet required for downloading model] ✅
• TTS Support ✅
• Microphone Support ✅
• HuggingFace Spaces Inference [for CPU Tier only] ✅
  • Remove Youtube & Input Path ✅
  • Remove Crepe Support due to gpu requirement ✅

Automatic Installation

Install ffmpeg first before running these command.

• Windows
  Run the start.bat to download the model and dependencies.
  Run the run.bat to run the inference
• MacOS & Linux
  For MacOS. before running the script, please install wget
  Run the start.sh to download the model and dependencies.
  Run the run.sh to run the inference

Manual Installation

1. Install Pytorch

   • CPU only (any OS)

   pip install torch torchvision torchaudio

   • Nvidia (CUDA used)

   # For Windows (Due to flashv2 not supported in windows, Issue: https://gith*ub.com**/Dao-AILab/flash-attention/issues/345#issuecomment-1747473481)
   pip install torch==2.1.0 torchvision==0.16.0 torchaudio==2.1.0 --index-url https://download.p**yto*rch.org/whl/cu121
   # Other (Linux, etc)
   pip install torch torchvision torchaudio --index-url https://download.p**yto*rch.org/whl/cu121

2. Install ffmpeg

3. Install Dependencies

pip install -r requirements.txt

4. Download Pre-model

# Hubert Model
https://**huggi*ngface.co/lj1995/VoiceConversionWebUI/blob/main/hubert_base.pt
# Save it to /assets/hubert/hubert_base.pt

# RVMPE (rmvpe pitch extraction, Optional)
https://hu*gg*i*ngface.co/lj1995/VoiceConversionWebUI/blob/main/rmvpe.pt
# Save it to /assets/rvmpe/rmvpe.pt

5. Run WebUI

python app.py

How to use

Command Line Arguments

Other Inference