Auto-Tune Model

This repository contains an auto-tune model for pitch correction using Python, Cog, and Replicate.

Quickstart

1. Install Cog

If you haven't already, install Cog:

pip install cog

2. Log in to Replicate

Authenticate your local Cog installation with Replicate:

cog login

Follow the instructions to paste your Replicate token.

3. Push the Model to Replicate

Push your model to Replicate (replace riffgen/auto-tune with your own repo if needed):

cog push r8.im/riffgen/auto-tune

4. Run the Model Locally

You can test the model locally by running:

python predict.py

Edit the __main__ block in predict.py to change the input parameters, for example:

if __name__ == "__main__":
    p = Predictor()
    p.setup()
    out = p.predict(
        audio_file="https://your-audio-url.m4a",
        autotune_style="balanced",
        scale="D:maj",
        correction_strength=0.8,
        adaptive_strength=True,
        smooth_transitions=True,
        plot=False,
        output_format="wav",
    )
    print("Output file:", out)

5. Run the Model on Replicate

After pushing, you can run the model on Replicate via the web UI or API. The UI will let you:

• Upload or link to an audio file
• Choose a musical scale (key)
• Choose an auto-tune style (intensity)
• Adjust correction strength, adaptive, and smoothing
• Download the processed audio

For API usage, see the Replicate Python client docs.

---

Questions?

• See Cog documentation
• See Replicate documentation
• Or open an issue in this repo!