Applio

// Overview

Applio is a voice conversion toolkit built on the RVC (retrieval-based voice conversion) approach, aimed at people who want high-quality results without assembling a pipeline themselves. Install scripts set up the environment, a launch script opens a Gradio interface in your browser, and from there you can convert audio with a trained voice model, prepare a dataset, train a new model and inspect the result. Its design is deliberately extensible — plugins and configuration let it be bent toward a particular project rather than locking you into one workflow.

Everything the UI does is also available from a click-based command line in core.py, which is what makes Applio scriptable. The commands cover single-file inference, batch inference over a folder, TTS synthesis followed by voice conversion, dataset preprocessing, feature extraction, training, index generation, model information, blending two models together, downloading a model from a link, installing prerequisites, and analysing an audio file. Inference takes an input path, an output path, and the .pth model plus its .index file.

The source code and the model weights in the repository are MIT-licensed, so modification, redistribution and commercial use are permitted. Using the official build as published also means accepting the project's Terms of Use, which govern its integrations, configurations and default behaviour; the maintainers ask that generated audio respect copyright, intellectual property and privacy rights. The project describes itself as stable and mature, with ongoing work focused on security patches, dependency updates and occasional feature improvements rather than frequent feature releases.

// What it does

• Gradio web interface launched by a single run script, plus a full click-based CLI in core.py
• Single-file and batch voice conversion from a trained .pth model and its .index file
• TTS synthesis piped through voice conversion, so text becomes speech in a target voice
• Full training path: dataset preprocessing, feature extraction, training, and index generation
• Model utilities — inspect a trained model, blend two models together, or download one from a link
• Plugin system, TensorBoard monitoring scripts, Dockerfiles, and ready-made Google Colab notebooks (UI and no-UI)

// Getting started

Applio is installed and launched by scripts checked into the repository — one to set up the environment, one to start the Gradio UI, one to attach TensorBoard. Colab notebooks are linked from the README if you would rather not install locally.

./run-install.sh        # Linux / macOS
# Windows: double-click run-install.bat

./run-applio.sh         # Linux / macOS
# Windows: double-click run-applio.bat

python core.py infer \
  --input-path /path/to/input.wav \
  --output-path /path/to/output.wav \
  --pth-path /path/to/model.pth \
  --index-path /path/to/model.index

python core.py preprocess --help
python core.py extract --help
python core.py train --help
python core.py index --help

./run-tensorboard.sh    # Linux / macOS
# Windows: run-tensorboard.bat

Commands and code are distilled from the project's own documentation — always check the official repo for the latest.

// When to use it

• Converting recorded audio into a trained target voice for dubbing, localisation or character work
• Training a voice model end to end from your own dataset, then monitoring the run in TensorBoard
• Batch-processing a folder of clips through one model from a script or CI job instead of the UI
• Turning a text file into speech in a specific voice by chaining TTS with voice conversion in one command

// How Applio compares

Applio alongside other open-source audio, music & voice tools AI/TLDR tracks, ranked by GitHub stars.