RVC-v2 + Beatrice-v2 Voice Conversion + training

RVC-v2 + Beatrice-v2 Voice Conversion + training

Converts a voice recording to another voice with an RVC v2 or Beatrice v2 model, and trains new models from your own audio, on a CPU-friendly Gradio page. It suits audio hobbyists and developers.

Made by Luminia on Hugging Face. These notes are Mindket's, written from its Space page and README; the app's own files and license are on Hugging Face.

About

What it is: This Space converts the voice in an audio recording into another voice, using either an RVC v2 model or a Beatrice v2 model. It can also train both kinds of model from your audio. The whole app sits in one file, runs on cpu-basic hardware at Hugging Face and is tagged MCP.

Who it suits: It suits people who work with voice models and want conversion and training on one page, and developers who want to call it from code or from an AI assistant.

Industry | Who uses it | How they use it | Example
Audio production | Hobbyist producers | Convert a recording with a voice model | Upload a take, pick RVC v2 and convert it
AI development | Developers | Call conversion and training from scripts | Use the Python client for batch work
Research | Voice model tinkerers | Train and test RVC or Beatrice models | Train a model from their own recordings

You can skip it if you want fast training, since training on a CPU works but is slow.

Where to use it: You can use it in these places:

• A browser: open the Space's page on Hugging Face and nothing needs installing.
• Windows, macOS or Linux: run your own copy with Python 3.10.
• Claude Desktop, Cursor and VS Code: through MCP, which exposes a convert tool and a train tool.
• Python scripts: through the gradio_client package.

Use case | Who | What they get back
Voice conversion | Producers | A converted audio file and a status message
Model training | Tinkerers | A trained model file and a status log
Batch processing | Developers | Converted files made through the command line or the Python client

How to set it up: You need these first for a local copy:

• Python 3.10 and pip
• A GPU, which the README recommends for training

The online page needs none of this. For your own copy, follow the steps below. The repository also has a requirements-training.txt file.

1. Install the Hugging Face tools.
2. Download the Space's files into a folder.
3. Move into that folder.
4. Install Gradio at the version the Space uses, then the other packages.
5. Start the app.

pip install -U huggingface_hub
hf download Luminia/rvc-beatrice-voice-conversion --repo-type space --local-dir rvc-beatrice-voice-conversion
cd rvc-beatrice-voice-conversion
pip install gradio==6.3.0
pip install -r requirements.txt
python app.py

How to use it: Open the Space in a browser. To convert, upload the source audio, choose RVC v2 or Beatrice v2 as the model type, upload the model file, adjust the pitch shift if needed and click Convert. To train, choose the trainer, upload training audio, enter a model name, set the epochs and batch size and click Start Training. From Python, the client call below runs a conversion with an RVC v2 model.

from gradio_client import Client, handle_file

client = Client("Luminia/rvc-beatrice-voice-conversion")

result = client.predict(
    source_audio=handle_file("voice.wav"),
    model_type="RVC v2",
    model_file=handle_file("model.pth"),
    index_file=None,
    beatrice_model_file=None,
    beatrice_target_speaker=0,
    beatrice_formant_shift=0.0,
    pitch_shift=0,
    f0_method="pm",
    index_rate=0.75,
    protect=0.33,
    api_name="/convert"
)
print(result)

Tool | What it does
convert | Converts source audio with an RVC v2 or Beatrice v2 model
train | Trains an RVC v2 or Beatrice v2 model from your audio
Command line | Handles batch processing

Example: You upload a voice recording, choose RVC v2 as the model type and upload a .pth model file. You click Convert and get back a converted audio file together with a status message.

For newcomers and experts: If you are new to this, try the default RVC v2 model with one short recording and leave the settings alone. Click Convert and listen to the result. Training asks for ten minutes or more of audio, so leave that for later.

If you know the field, the app reads RVC v1 and v2 checkpoints, with or without f0, and detects the version from the checkpoint. It supports index retrieval files for RVC. Beatrice v2 takes 16 kHz input, gives 24 kHz output and uses a per-speaker VQ.

Upload source audio
Choose RVC v2 or Beatrice v2
Upload the model file
Set the pitch shift
Click Convert and get audio

Good to know: The hosted Space handles only a single job at once, so jobs wait in line. A copy on your own computer avoids the wait.

• License: mit
• Cost to run: none named
• Data it reaches: the audio and model files you upload
• Not for: quick training on a CPU, which works but is slow

On Hugging Face, 13 November 2026

• Likes: 42
• Built with: gradio
• License: mit

Links

• Space: https://huggingface.co/spaces/Luminia/rvc-beatrice-voice-conversion
• Files: https://huggingface.co/spaces/Luminia/rvc-beatrice-voice-conversion/tree/main

Try it

Open the Space in your browser, or copy it to your own Hugging Face account from its page. Nothing in this kit runs by itself.