Using the Voice Blender

The Voice Blender is a unique feature in Applio that allows you to fuse two existing voice models into a single new model. This can be a fun way to experiment and create new, unique voices.

How to Use the Voice Blender

1. Go to the Voice Blender tab in the Applio interface.
2. Enter a Name: Give your new, blended model a unique name.
3. Select Model A: Drag and drop the .pth file for your first model into the “Model A” box.
4. Select Model B: Drag and drop the .pth file for your second model into the “Model B” box.
5. Adjust the Blend Ratio: Use the slider to adjust the blend ratio between the two models. A value of 0.5 will create an even blend, while a value closer to 0 will favor Model A, and a value closer to 1 will favor Model B.
6. Fuse: Click the Fusion button to create your new model.

Your new blended model will be saved in the Applio/logs directory and will be available for use in the Inference tab.

Note

Blended models do not have an index file, so you will not need to select one when using them for inference.

Tips for Best Results

• Experiment: The best way to get interesting results with the Voice Blender is to experiment with different combinations of models and blend ratios.
• Use Similar Models: You’ll often get the best results when blending models that have similar vocal characteristics.
• Start with a 50/50 Blend: A good starting point is to set the blend ratio to 0.5 and then adjust from there.