Zyphra's Zonos-v0.1 Challenges ElevenLabs With a Free Open-Weight Voice Model

Zonos-v0.1 is Zyphra's open-weight text-to-speech model, and it's making a strong case that proprietary TTS APIs no longer have a quality moat. Zyphra is releasing both a 1.6B transformer and a 1.6B hybrid model under an Apache 2.0 license, meaning you can use them commercially, self-host them, and redistribute them without restrictions. With nearly 10,000 downloads and close to 3,000 likes on Hugging Face, the community has taken notice.

Zonos-v0.1 is trained on more than 200,000 hours of varied multilingual speech, delivering expressiveness and quality on par with , or even surpassing , top TTS providers. That's a bold claim, but Zyphra backs it up with direct audio comparisons against ElevenLabs, Cartesia, and Fish Speech on their official blog.

Two models, one suite

The suite comprises a transformer model and an SSM hybrid model , notably, the first open-source SSM (State Space Model) model available for TTS. SSMs are a class of sequence models that process data more efficiently than pure transformers by avoiding the quadratic cost of full attention. This dual approach allows Zyphra to thoroughly investigate the performance and quality tradeoffs between these architectures in audio generation.

In practice, the two models serve slightly different needs. The transformer variant is the safer pick for compatibility , it runs on a wider range of hardware. The hybrid model trades some compatibility for reduced latency and memory overhead, thanks to its Mamba2-based architecture that relies less heavily on attention blocks.

What it can do

The model enables highly natural speech generation from text prompts when given a speaker embedding or audio prefix, and can accurately perform speech cloning when given a reference clip spanning just a few seconds. The conditioning setup also allows for fine control over speaking rate, pitch variation, audio quality, and emotions such as happiness, fear, sadness, and anger.