MockingBird
Clone a voice in 5 seconds to generate arbitrary speech in real-time.
Clone voices in 5 seconds and generate custom speech in real-time using PyTorch.
MockingBird is a real-time voice cloning and text-to-speech toolkit built on PyTorch. It suits civil engineering teams and technical professionals needing rapid localized audio generation for project presentations or site warnings.
What it is
It processes audio datasets and mel spectrograms to train custom encoders, synthesizers, and vocoders. You can reuse pretrained encoders while adapting the synthesizer for specific Mandarin dialects or multi-dataset inputs.
Who it's for
It suits Python developers and technical specialists with deep learning experience. It does not suit non-technical users looking for a simple out-of-the-box text-to-speech mobile app.
Key features
5-Second Voice Cloning
Clone a target voice using just a 5-second audio sample.
PyTorch 1.9.0 Support
Built and tested using PyTorch with GPU support like Tesla T4 and GTX 2060.
Webserver Ready
Serve your generated speech results easily with remote calling endpoints.
Multi-Dataset Training
Preprocess and train with Chinese datasets like aidatatang_200zh, magicdata, and aishell3.
Up and running in 3 steps
Our honest verdict
What's great
- Clones arbitrary voices from very short 5-second audio samples
- Supports webserver deployment for remote API calling
- Includes specific setup documentation for M1 Macs using Rosetta
- Reuses pretrained encoders and vocoders to simplify training
Where it falls short
- Requires manual compilation of dependencies like pyworld and ctc-segmentation on M1 Macs
- Training custom synthesizers requires downloading external datasets and managing checkpoints
- Original CLI tool is incompatible with Chinese symbols, requiring custom model paths
Pricing & license
Free under its MIT licence. Additionally, the original author offers a cloud-hosted optimized version called noiz.ai.
Open remote AI-training roles
766 open remote AI-training roles, updated November 14, 2026; most pay $20–$175 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
Use it, or have us set it up
FAQ
Does this tool support English?
The repository mentions English support alongside Mandarin, though testing heavily focuses on Chinese datasets and models.
Can I run this on an Apple Silicon Mac?
Yes, but you must follow specific workarounds involving Rosetta 2 terminal setups and manual source compilation for certain packages.
