babysor/MockingBird
🔧 Toolbabysor
Clone a voice in 5 seconds to generate arbitrary speech in real-time.
MockingBird is an open-source voice cloning system that allows users to clone any voice from a short audio snippet (as little as 5 seconds) and then generate arbitrary speech in that voice in real-time. The project uses a deep learning architecture combining a speaker encoder, a text-to-speech (TTS) synthesizer (like Tacotron2), and a neural vocoder (e.g., WaveGlow or HiFi-GAN). The speaker encoder extracts a speaker embedding from the short sample, which conditions the TTS model to produce speech that mimics the original speaker's characteristics. The system is optimized for low latency and can run on consumer GPUs. MockingBird also provides pretrained models and scripts for training custom voices. With over 36,000 GitHub stars, it is one of the most popular voice cloning projects, demonstrating significant interest and community contributions.
💡Highlights
- ├─Clone any voice in 5 seconds
- ├─Real-time arbitrary speech generation
- └─Built with PyTorch
🎯For
- ├─AI researchers
- ├─developers
- └─content creators