All models
voice-cloning

Llasa

CC-BY-NC-4.0

LLM-native TTS built straight on the Llama architecture

Llasa from HKUST Audio is an LLM-native TTS model — it runs the unmodified Llama architecture over XCodec2 speech tokens instead of bolting on a separate audio decoder, and does zero-shot voice cloning from a short reference. It ships in 1B, 3B, and 8B variants, trading size for fidelity.

Hear Llasa

Llasa-3B · kjjk10/llasa-3b-long · synthetic Bella reference · Same three scripts across the directory.

Run it yourself

Best for: LLM-native TTS built straight on the Llama architecture

Noncommercial / separate license: The official card explicitly prohibits free commercial use. Its Llama and codec dependencies do not override that restriction.

Code: Check the selected inference recipe separately. Weights: CC-BY-NC-4.0; Llama base also identified. License sources checked 2026-09-20. Source 1

Before you choose: Compare the samples with your own use case. Check the linked documentation for code, weight, and voice license terms.

Details awaiting a fresh review. Hardware figures are estimates unless a benchmark is linked.

Compare with similar

Compare all →
F5-TTSListen side-by-side →OpenVoice v2Listen side-by-side →Chatterbox TurboListen side-by-side →

Choose an alternative to hear both models together.

Voices

1 voice
Female
Bella referencedefault
American

The scripts

Neutral
“The quick brown fox jumps over the lazy dog near the riverbank.”
Emotional
“I can't believe you actually did it. This is incredible!”
Numbers & Dates
“On March 14th, 2025, the team raised $4.2 million at a 38% margin.”

Run Llasa

git clone https://huggingface.co/HKUSTAudio/Llasa-3B

Sources & setup details

Llasa FAQ

Do I need a GPU to run Llasa?

The directory lists roughly 12 GB of VRAM as a historical estimate. Check the model documentation for your checkpoint and runtime.

Can Llasa clone voices?

Yes — Llasa supports voice cloning from reference audio.

What license is Llasa released under?

Llasa is released under the CC-BY-NC-4.0 license. Always check the repository for the exact terms — some models license code and weights separately.

What languages does Llasa support?

Llasa supports English, Chinese.

Is there a hosted Llasa API?

OpenSpeech Cloud is accepting production-access interest for Llasa, but this model is not currently hosted in Cloud. Use the official repository for current deployment options.