Spr-Aachen/Easy-Voice-Toolkit

Easy Voice Toolkit – Quick Overview

What it is – A desktop‑plus‑Colab toolkit that strings together a handful of open‑source voice‑related models (e.g., Whisper for transcription, GPT‑SoVITS for voice conversion) and utilities so you can take raw audio, clean it, transcribe it, build a dataset, train a voice‑conversion model, and finally generate converted speech. The project ships a ready‑to‑run Windows portable package and also provides a Jupyter notebook for cloud‑based use.

Key capabilities

• Audio processing – trimming, slicing, and format handling.
• Voice recognition & transcription – powered by OpenAI’s Whisper.
• Dataset creation – tools to prepare paired audio for voice‑conversion training.
• Model training – scripts that wrap the GPT‑SoVITS pipeline for training a new voice model.
• Voice conversion – convert an input speaker’s voice into the style of a trained target voice.

How to get started

1. Windows users – click the Download Windows Portable Package badge; unzip and run run.py (no Python install needed).
2. Colab users – open the Google Colab Demo notebook via the badge, follow the cells to install dependencies and run the same workflow in the cloud.
3. Developers – clone the repo (with submodules), install Python ≥ 3.8, install PyTorch matching your CUDA version, then pip install -r requirements.txt and pip install QEasyWidgets. Run run.py to launch the GUI client and backend server.

Typical workflow

1. Import raw audio → optional cleaning/slicing.
2. Transcribe with Whisper to get a text transcript.
3. Create a paired dataset (source audio + target voice samples).
4. Train a GPT‑SoVITS voice‑conversion model on that dataset.
5. Convert new utterances, producing speech that sounds like the target voice.

Legal / usage notes

• Intended for academic and learning purposes only; not for production.
• When publishing results (e.g., videos), you must disclose the source audio and give credit to the original creators and this repository.
• The authors ask that you contact them before using the toolkit for any commercial or non‑research plan.

Roadmap

• Planned integration of a chatbot (LLM) for interactive voice‑assistant features.
• Refactoring the client UI to C++/Qt for better performance.
• Ongoing backend work and upcoming Linux support.

Acknowledged upstream projects – audio‑slicer, VoiceprintRecognition‑Pytorch, OpenAI Whisper, SRT‑to‑CSV‑and‑audio‑split, and GPT‑SoVITS.

---

All information above is taken directly from the repository’s README; no additional features have been inferred.