Text-to-Speech
Transformers
Safetensors
higgs_multimodal_qwen3
text-generation
tts
voice-cloning
multilingual
expressive-speech
controllable-tts
higgs-audio
mirror
Instructions to use AEmotionStudio/higgs-tts-3-models with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use AEmotionStudio/higgs-tts-3-models with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-to-speech", model="AEmotionStudio/higgs-tts-3-models")# pip install -U transformers accelerate # Load model directly from transformers import AutoModelForSeq2SeqLM model = AutoModelForSeq2SeqLM.from_pretrained("AEmotionStudio/higgs-tts-3-models", device_map="auto") - Notebooks
- Google Colab
- Kaggle
Higgs TTS 3 (4B) — mirror for MAESTRO
Unmodified, sha256-verified mirror of bosonai/higgs-tts-3-4b (Boson AI), kept by AEmotionStudio so the MAESTRO DAW can fetch the checkpoint on demand. Nothing here is fine-tuned, quantised or re-packed.
| File | Purpose |
|---|---|
model.safetensors (+ .index.json) |
4B talker (Qwen3-4B backbone + fused 8-codebook audio head) and the bundled Higgs audio tokenizer (tied.embedding.modality_embeddings.0.model.*) |
config.json |
HiggsMultimodalQwen3ForConditionalGeneration config |
tokenizer.json, tokenizer_config.json |
Qwen2 text tokenizer with the `< |
PROMPTING.md |
Boson's control-tag placement guide |
LICENSE, NOTICE |
The governing license and the attribution notice required by its §IV(a) |
License — please read
Boson Higgs TTS 3 Research and Non-Commercial License (Boson AI USA, Inc.).
- Research and personal / hobbyist use: permitted.
- Creator Use Grant: digital creators may create, publish and monetise podcasts, videos, audiobooks and social posts made with the model, provided they credit "Boson AI's Higgs Audio" (in the audio or in the accompanying text, e.g. "This audio was created with Boson AI's Higgs Audio — https://www.boson.ai/higgs-audio").
- Hosting the model as a service, embedding it in a product or application made available to third parties, or reselling / redistributing derivatives requires a separate commercial license from Boson AI (contact@boson.ai).
- Prohibited: cloning or impersonating a real person's voice without their explicit, verifiable consent; fraud or deception; election-related deception; biometric surveillance; training other generative models on the outputs.
By downloading these files you accept the LICENSE in this repository. MAESTRO shows the same terms in-app and asks for acknowledgement before the first generation.
Citation
@misc{bosonai_higgs_audio_tts_v3_2026,
title = {Higgs TTS 3: Conversational Speech for Voice AI from Boson AI},
author = {Boson AI},
year = {2026},
howpublished = {https://huggingface.co/bosonai/higgs-tts-3-4b},
}
- Downloads last month
- 8