voicepaw/so-vits-svc-fork
So Vits Svc Fork
This project is an AI singing voice conversion system and vocal processor used for training generative voice models and converting vocal recordings or live input into a target voice. It functions as a VITS model trainer and a real-time voice changer that transforms vocal timbre and pitch to change the identity of a singer.
The system provides a graphical management dashboard for controlling training hyperparameters and voice conversion presets. It supports low-latency audio streaming for live microphone input and employs pitch estimation to ensure precise matching between source and target voice tones while preserving original melody and lyrics.
The software covers the full pipeline from data preparation to inference, including audio dataset preprocessing, recording segmentation, and voice model training with visual progress monitoring. It also includes utilities for batch audio processing and hardware routing for managing physical audio input and output devices.
Features
- Real-Time Voice Transformation - Provides a low-latency system for modifying live microphone input into a target voice using machine learning.
- Singing Voice Conversions - Transforms vocal characteristics of recordings to match a target voice while preserving original melody and lyrics.
- AI Vocal Production - Acts as a vocal processor for pitch estimation and timbre transformation to change a singer's identity.
- Voice Model Trainers - Ships a specialized tool for training and fine-tuning generative voice models based on the VITS architecture.
- Training Hyperparameters - Allows fine-tuning of learning rates and gradient accumulation to optimize the model's ability to mimic specific voices.
- Voice Synthesizer Training - Implements a full pipeline for preparing audio datasets and training neural networks to replicate specific voices.
- Voice Identity Conversions - Transforms source audio spectral envelopes and fundamental frequency to map a voice onto a target speaker identity.
- Singing - Specializes in transforming the vocal characteristics of singing recordings to a target voice while preserving melody.
- Singing Voice Conversion Systems - A full system for training generative voice models and converting vocal recordings or live input using VITS.
- Pitch Estimation - Analyzes audio frequencies to estimate the fundamental frequency for precise matching between source and target tones.
- Audio Segmenting - Segments long audio recordings into smaller clips based on silence or speaker changes to prepare training data.
- Audio Dataset Preprocessing - Provides tools for cleaning, segmenting, and standardizing raw audio recordings for ML training.
- Management Dashboards - Provides a graphical management dashboard for controlling training hyperparameters and voice conversion presets.
- Training Dataset Preparation - Cleans and structures raw audio recordings into a format suitable for machine learning training.
- Batch Processing - Supports automated workflows for running voice conversion inference across multiple audio files simultaneously.
- Low-Latency Processing Buffers - Processes audio in small buffer chunks to enable real-time voice conversion between hardware input and output.
- Conversion Control Interfaces - Offers a graphical interface to manage voice conversion tasks and settings without requiring the command line.
