AI product Open source
tiny-voca is an open English text-to-speech model developed by Nazeem Muhammed. The 4.5-million-parameter model generates 24 kHz American English speech in two fixed voices, F2 and M2, and includes its waveform generator. It runs with PyTorch, ONNX Runtime, or ONNX Runtime Web, making it usable on CPUs and in browsers; the repository reports about 70× real-time generation on four CPU threads. The project includes its code, weights, training recipe, and dataset, with command-line and Python interfaces, sentence-by-sentence handling for long text, adjustable speaking rate and pitch, and a browser demo. It is limited to English and two voices and does not support voice cloning or streaming. Code and weights are licensed under Apache-2.0, while the dataset is licensed under CC BY 4.0.
1 use taken from transcripts — each links to the moment in the video.
A text-to-speech model with just 4.5 million parameters trained from scratch in PyTorch, small enough to run on a CPU or in the browser; it speaks American English in two fixed voices at 24 kHz and runs about 70x real time on four CPU threads.
1 in the library.