higgs-tts-3-4b

Kremon96/higgs-tts-3-4b

higgs-tts-3-4b is a text-to-speech model from Kremon96. Use it when you need text read aloud. It is set up for transformers. The card lists the license as other.

<div align="left" style="display: flex; justify-content: flex-start; margin-top: 10px;" <a href="https://www.boson.ai/blog/higgs-audio-v3-tts" <img src="https://img.shields.io/badge/💚-Boson%20Blog-228B22" style="marg…

Downloads · 30 days

38% of all-time downloads

All-time downloads

Public

Parameters

4.7B

9.3 GB on disk

Likes

Public

Hugging Face

Repo makeup

Click a slice to open those files.

.safetensors9.3 GB · 100%

At a glance

Task: Text-to-Speech
Library: transformers
License: other
Model type: higgs_multimodal_qwen3
Access: Public
Created: Sep 24, 2026
Updated: Sep 24, 2026
SHA: f73b58dc

Model card

From the Hugging Face model README

Higgs TTS 3

Higgs TTS 3 is built for voice chat: it speaks, not just reads. It turns model responses into expressive conversational speech across 100+ languages, with zero-shot voice cloning and inline control over emotion, style, prosody, pauses, and sound effects.

[!TIP] Released for research and non-commercial use under the Boson Higgs TTS 3 Research and Non-Commercial License. Production, hosted APIs, embedding in a product/service, or reselling the model requires a separate commercial license. Prohibited: voice cloning without consent, impersonation, fraud, election deception, biometric surveillance, or any unlawful use.

[!TIP] Free for digital creators — including monetized content. Under the license's Creator Use Grant, creators may use Higgs TTS 3 to make and monetize podcasts, videos, and social posts for free. The one requirement is to credit Boson AI's Higgs Audio — either in the audio or prominently in the accompanying text (e.g., the video description or show notes). Suggested credit: "This audio was created with Boson AI's Higgs Audio — https://www.boson.ai/higgs-audio". See the Creator Use section below.

Higgs TTS 3 Architecture

Higgs autoregressive decoder consumes interleaved text and audio tokens. Audio is encoded by the Higgs Tokenizer into 8 codebooks at 25 fps, staggered via a delay pattern, then mapped to backbone hidden states through a multi-codebook fused embedding. Output codes pass through a multi-codebook fused head, are de-delayed, and decoded back to waveform.

Component	Spec
Backbone	~4B autoregressive decoder (36 L, hidden=2560, GQA 32/8)
Multi-codebook embedding / head	Fused single-tensor, tied with text embedding
Context length	8,192 tokens (training sequence length)
Audio tokens	8 codebooks × 1026 vocab, delay pattern
Sample rate	24 kHz
Frame rate	25 fps (40 ms / frame)

Supported Languages

The model reaches single-digit WER/CER on 102 languages, which split into two tiers.

WER/CER under 5 — polished, production-quality (85)

🇿🇦 Afrikaans · 🇸🇦🇪🇬 Arabic · 🇦🇲 Armenian · 🇮🇳 Assamese · 🇪🇸 Asturian · 🇦🇿 Azerbaijani · 🇷🇺 Bashkir · 🇪🇸 Basque · 🇧🇾 Belarusian · 🇧🇩🇮🇳 Bengali · 🇧🇦 Bosnian · 🇧🇬 Bulgarian · 🇪🇸 Catalan · 🇵🇭 Cebuano · 🇮🇶 Central Kurdish · 🇨🇳 Chinese · 🇭🇷 Croatian · 🇨🇿 Czech · 🇩🇰 Danish · 🇳🇱🇧🇪 Dutch · 🇷🇺 Eastern Mari · 🇺🇸🇬🇧🇦🇺 English · 🌐 Esperanto · 🇪🇪 Estonian · 🇫🇮 Finnish · 🇫🇷🇨🇦 French · 🇪🇸 Galician · 🇬🇪 Georgian · 🇩🇪🇦🇹 German · 🇬🇷 Greek · 🇮🇳 Gujarati · 🇭🇹 Haitian Creole · 🇳🇬 Hausa · 🇮🇱 Hebrew · 🇮🇳 Hindi · 🇭🇺 Hungarian · 🇮🇩 Indonesian · 🇮🇹 Italian · 🇯🇵 Japanese · 🇮🇩 Javanese · 🇮🇳 Kannada · 🇰🇿 Kazakh · 🇰🇷 Korean · 🇷🇼 Kinyarwanda · 🇰🇬 Kyrgyz · 🇱🇻 Latvian · 🇨🇩 Lingala · 🇱🇹 Lithuanian · 🇰🇪 Luo · 🇲🇰 Macedonian · 🇲🇾🇮🇩 Malay · 🇮🇳 Malayalam · 🇲🇹 Maltese · 🇳🇿 Māori · 🇮🇳 Marathi · 🇲🇳 Mongolian · 🇳🇵 Nepali · 🇳🇴 Norwegian · 🇫🇷 Occitan · 🇮🇷🇦🇫 Persian · 🇵🇱 Polish · 🇵🇹🇧🇷 Portuguese · 🇷🇴 Romanian · 🇷🇺 Russian · 🇿🇦 Sepedi · 🇷🇸 Serbian · 🇿🇼 Shona · 🇸🇰 Slovak · 🇸🇮 Slovene · 🇪🇸🇲🇽 Spanish · 🇹🇿🇰🇪 Swahili · 🇸🇪 Swedish · 🇵🇭 Tagalog · 🇹🇯 Tajik · 🇮🇳🇱🇰 Tamil · 🇮🇳 Telugu · 🇹🇭 Thai · 🇹🇷 Turkish · 🇺🇦 Ukrainian · 🇵🇰🇮🇳 Urdu · 🇨🇳 Uyghur · 🇺🇿 Uzbek · 🇻🇳 Vietnamese · 🇿🇦 Xhosa · 🇿🇦 Zulu

WER/CER between 5 and 10 — usable, but less polished (17)

🇦🇱 Albanian · 🇲🇼🇿🇲 Chichewa/Nyanja · 🇮🇳🇵🇰 Eastern Punjabi · 🇺🇬 Ganda · 🇮🇸 Icelandic · 🇮🇪 Irish · 🇩🇿 Kabyle · 🇨🇻 Kabuverdianu · 🇰🇪 Kamba · 🇻🇦 Latin · 🇱🇺 Luxembourgish · 🇪🇹🇰🇪 Oromo · 🇦🇫🇵🇰 Pashto · 🇵🇰🇮🇳 Sindhi · 🇸🇴 Somali · 🇦🇴 Umbundu · 🇬🇧 Welsh

Control Tokens

All tags follow <|category:value|> syntax and can be inserted mid-utterance.

For how to place these tags when writing the target text (sentence-level vs. inline, sfx formatting, stacking, worked examples), see PROMPTING.md.

Emotion — elation, amusement, enthusiasm, determination, pride, contentment, affection, relief, contemplation, confusion, surprise, awe, longing, arousal, anger, fear, disgust, bitterness, sadness, shame, helplessness

<div style="margin-left:1.5em;margin-top:-10px"> <table style="display:inline-table;width:max-content;max-width:none;border-collapse:collapse;font-size:clamp(11px,1.4vw,13px);text-align:left"> <thead><tr> <th style="padding:14px 20px;text-align:left;font-weight:600;border-bottom:2px solid #7BCFA3;color:#7BCFA3">Token</th> <th style="padding:14px 20px;text-align:left;font-weight:500;border-bottom:2px solid #7BCFA3;color:#7BCFA3">Description</th> </tr></thead> <tbody> <tr><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)"><code><|emotion:elation|></code></td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Elation / joy</td></tr> <tr><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)"><code><|emotion:amusement|></code></td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Amusement / playful laughter</td></tr> <tr><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)"><code><|emotion:enthusiasm|></code></td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Enthusiasm / excitement</td></tr> <tr><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)"><code><|emotion:determination|></code></td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Determination / firmness</td></tr> <tr><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)"><code><|emotion:pride|></code></td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Pride / confidence</td></tr> <tr><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)"><code><|emotion:contentment|></code></td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Calm satisfaction</td></tr> <tr><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)"><code><|emotion:affection|></code></td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Warmth / affection</td></tr> <tr><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)"><code><|emotion:relief|></code></td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Relief</td></tr> <tr><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)"><code><|emotion:contemplation|></code></td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Thoughtful / reflective</td></tr> <tr><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)"><code><|emotion:confusion|></code></td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Confused</td></tr> <tr><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)"><code><|emotion:surprise|></code></td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Surprised</td></tr> <tr><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)"><code><|emotion:awe|></code></td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Awe / wonder</td></tr> <tr><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)"><code><|emotion:longing|></code></td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Longing / yearning</td></tr> <tr><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)"><code><|emotion:arousal|></code></td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Heightened desire</td></tr> <tr><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)"><code><|emotion:anger|></code></td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Anger</td></tr> <tr><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)"><code><|emotion:fear|></code></td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Fear</td></tr> <tr><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)"><code><|emotion:disgust|></code></td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Disgust</td></tr> <tr><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)"><code><|emotion:bitterness|></code></td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Bitterness</td></tr> <tr><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)"><code><|emotion:sadness|></code></td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Sadness</td></tr> <tr><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)"><code><|emotion:shame|></code></td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Shame</td></tr> <tr><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)"><code><|emotion:helplessness|></code></td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Helplessness</td></tr> </tbody> </table> </div>

Style — singing, shouting, whispering

Sound effects — cough, laughter, crying, screaming, burping, humming, sigh, sniff, sneeze

<div style="margin-left:1.5em;margin-top:-10px"> <p style="margin:0 0 1px;color:rgba(128,128,128,0.9);font-size:13px;text-align:left">Pair each token with the matching onomatopoeia immediately after it.</p> <table style="display:inline-table;width:max-content;max-width:none;border-collapse:collapse;font-size:clamp(11px,1.4vw,13px);text-align:left"> <thead><tr> <th style="padding:14px 20px;text-align:left;font-weight:600;border-bottom:2px solid #7BCFA3;color:#7BCFA3">Token</th> <th style="padding:14px 20px;text-align:left;font-weight:500;border-bottom:2px solid #7BCFA3;color:#7BCFA3">Description</th> <th style="padding:14px 20px;text-align:left;font-weight:500;border-bottom:2px solid #7BCFA3;color:#7BCFA3">Suggested onomatopoeia</th> </tr></thead> <tbody> <tr><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)"><code><|sfx:cough|></code></td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Cough</td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Ahem</td></tr> <tr><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)"><code><|sfx:laughter|></code></td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Laughter</td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Haha / Hehe</td></tr> <tr><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)"><code><|sfx:crying|></code></td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Crying</td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Boohoo / Sob</td></tr> <tr><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)"><code><|sfx:screaming|></code></td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Screaming</td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Ahh / Aaah</td></tr> <tr><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)"><code><|sfx:burping|></code></td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Burping</td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Burp</td></tr> <tr><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)"><code><|sfx:humming|></code></td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Humming</td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Hmm / Mmm</td></tr> <tr><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)"><code><|sfx:sigh|></code></td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Sigh</td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Uh / Ahh</td></tr> <tr><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)"><code><|sfx:sniff|></code></td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Sniff</td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Sff</td></tr> <tr><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)"><code><|sfx:sneeze|></code></td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Sneeze</td><td style="padding:12px 20px;border-bottom:1px solid rgba(128,128,128,0.15)">Achoo</td></tr> </tbody> </table> </div>

Prosody
- Speed — speed_very_slow, speed_slow, speed_fast, speed_very_fast
- Pauses — pause, long_pause
- Pitch — pitch_low, pitch_high
- Delivery — expressive_high, expressive_low

Evaluation Benchmarks

Multilingual Voice Clone

We evaluate Higgs TTS 3 on public multilingual TTS suites and our internal 111-language Higgs-Multilingual set, covering both common and lower-resource languages.

WER / CER (↓, ×100) macro-averaged across each benchmark's language set. Lower is better; bold marks the best per row. All numbers are reproducible end-to-end with original metrics and normalization.

<div style="font-family:-apple-system,BlinkMacSystemFont,'Segoe UI',Roboto,sans-serif;width:100%;padding:16px 0;overflow-x:auto;-webkit-overflow-scrolling:touch"> <table style="width:100%;border-collapse:collapse;font-size:clamp(11px,1.4vw,13px);text-align:center"> <thead><tr> <th style="padding:18px 24px;text-align:center;font-weight:600;border-bottom:2px solid #16a34a;color:#16a34a">Benchmark</th> <th style="padding:18px 24px;text-align:center;font-weight:500;white-space:nowrap;border-bottom:2px solid #16a34a;color:#16a34a;font-size:14px">Higgs TTS v2</th> <th style="padding:18px 24px;text-align:center;font-weight:500;white-space:nowrap;border-bottom:2px solid #16a34a;color:#16a34a;font-size:14px">Higgs TTS 3</th> <th style="padding:18px 24px;text-align:center;font-weight:500;white-space:nowrap;border-bottom:2px solid #16a34a;color:#16a34a;font-size:14px">Fish Audio S2 Pro</th> <th style="padding:18px 24px;text-align:center;font-weight:500;white-space:nowrap;border-bottom:2px solid #16a34a;color:#16a34a;font-size:14px">Qwen3-TTS-1.7B</th> <th style="padding:18px 24px;text-align:center;font-weight:500;white-space:nowrap;border-bottom:2px solid #16a34a;color:#16a34a;font-size:14px">VibeVoice-7B</th> <th style="padding:18px 24px;text-align:center;font-weight:500;white-space:nowrap;border-bottom:2px solid #16a34a;color:#16a34a;font-size:14px">IndexTTS-2</th> <th style="padding:18px 24px;text-align:center;font-weight:500;white-space:nowrap;border-bottom:2px solid #16a34a;color:#16a34a;font-size:14px">MiMo-Audio-7B-Instruct</th> <th style="padding:18px 24px;text-align:center;font-weight:500;white-space:nowrap;border-bottom:2px solid #16a34a;color:#16a34a;font-size:14px">MOSS-TTS-v1.5</th> <th style="padding:18px 24px;text-align:center;font-weight:500;white-space:nowrap;border-bottom:2px solid #16a34a;color:#16a34a;font-size:14px">OmniVoice</th> <th style="padding:18px 24px;text-align:center;font-weight:500;white-space:nowrap;border-bottom:2px solid #16a34a;color:#16a34a;font-size:14px">ChatterBox</th> <th style="padding:18px 24px;text-align:center;font-weight:500;white-space:nowrap;border-bottom:2px solid #16a34a;color:#16a34a;font-size:14px">FireRedTTS-2</th> </tr></thead> <tbody> <tr> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">SeedTTS</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">2.10</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-weight:600">1.11</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">1.31</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">1.30</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">3.59</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">1.63</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">3.70</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">1.73</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">1.21</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">17.00</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">1.72</td> </tr> <tr> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">CV3</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">21.19</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-weight:600">4.41</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">4.60</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">7.73</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">11.66</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">129.26</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">71.55</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">6.11</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">4.92</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">32.62</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">19.20</td> </tr> <tr> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">MiniMax-Multilingual</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">49.86</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-weight:600">2.74</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">5.15</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">27.41</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">8.21</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">112.91</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">85.67</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">3.78</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">2.98</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">49.30</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">12.52</td> </tr> <tr> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">Higgs-Multilingual</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">52.24</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-weight:600">3.61</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">8.68</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">97.09</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">13.74</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">57.71</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">59.61</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">21.28</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">3.63</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">57.52</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">33.69</td> </tr> </tbody> </table> </div>

Emergent TTS

Win-rate (↑) per category — judge preference vs the BASELINE row; bold marks the highest win-rate per column. For a fair comparison, every model shares the same reference audio per prompt, and we run the benchmark text verbatim — no inline control tags inserted.

<div style="font-family:-apple-system,BlinkMacSystemFont,'Segoe UI',Roboto,sans-serif;width:100%;padding:16px 0;overflow-x:auto;-webkit-overflow-scrolling:touch"> <table style="width:100%;border-collapse:collapse;font-size:clamp(11px,1.4vw,13px);text-align:center"> <thead><tr> <th style="padding:18px 24px;text-align:center;font-weight:600;border-bottom:2px solid #16a34a;color:#16a34a">Model</th> <th style="padding:18px 24px;text-align:center;font-weight:500;white-space:nowrap;border-bottom:2px solid #16a34a;color:#16a34a;font-size:14px">Overall ↑</th> <th style="padding:18px 24px;text-align:center;font-weight:500;white-space:nowrap;border-bottom:2px solid #16a34a;color:#16a34a;font-size:14px">Emotions ↑</th> <th style="padding:18px 24px;text-align:center;font-weight:500;white-space:nowrap;border-bottom:2px solid #16a34a;color:#16a34a;font-size:14px">Foreign Words ↑</th> <th style="padding:18px 24px;text-align:center;font-weight:500;white-space:nowrap;border-bottom:2px solid #16a34a;color:#16a34a;font-size:14px">Paralinguistics ↑</th> <th style="padding:18px 24px;text-align:center;font-weight:500;white-space:nowrap;border-bottom:2px solid #16a34a;color:#16a34a;font-size:14px">Complex Pronunciation ↑</th> <th style="padding:18px 24px;text-align:center;font-weight:500;white-space:nowrap;border-bottom:2px solid #16a34a;color:#16a34a;font-size:14px">Questions ↑</th> <th style="padding:18px 24px;text-align:center;font-weight:500;white-space:nowrap;border-bottom:2px solid #16a34a;color:#16a34a;font-size:14px">Syntactic Complexity ↑</th> </tr></thead> <tbody> <tr> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">Higgs TTS 3</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-weight:600">53.65%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">53.75%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-weight:600">48.75%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-weight:600">68.57%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">25.10%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-weight:600">61.43%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-weight:600">60.71%</td> </tr> <tr> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">Fish Audio S2 Pro</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">43.80%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">53.04%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">33.93%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">53.75%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">18.16%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">55.00%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">45.71%</td> </tr> <tr> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">Qwen3-TTS-1.7B</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">38.84%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">45.54%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">24.64%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">44.29%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-weight:600">30.00%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">53.39%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">34.11%</td> </tr> <tr> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">IndexTTS-2</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">31.12%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">39.29%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">5.36%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">42.50%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">12.45%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">45.89%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">38.93%</td> </tr> <tr> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">MOSS-TTS-v1.5</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">43.89%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">60.54%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">35.18%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">51.43%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">11.63%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">53.21%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">47.32%</td> </tr> <tr> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">OmniVoice</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">40.82%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15);font-weight:600">61.07%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">28.75%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">52.68%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">13.67%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">45.00%</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">40.36%</td> </tr> </tbody> </table> </div>

Usage

SGLang Usage

Pair the weights in this repo with SGLang-Omni — a production serving stack with continuous batching for multi-codebook decoding and the same inline tag controls. The Higgs TTS cookbook walks you through installation, server launch, request examples, and the full API reference.

See the Higgs TTS cookbook for the full details.

Install and Serve

docker pull lmsysorg/sglang-omni:dev
docker run -it --gpus all --shm-size 32g --ipc host --network host --privileged \
  lmsysorg/sglang-omni:dev /bin/zsh

git clone [email protected]:sgl-project/sglang-omni.git && cd sglang-omni
uv venv .venv -p 3.12 && source .venv/bin/activate
uv pip install -v -e .

export HF_TOKEN=hf_xxxxxxxxxxxxxxxx
hf download bosonai/higgs-tts-3-4b

sgl-omni serve \
  --model-path bosonai/higgs-tts-3-4b \
  --port 8000

Zero-shot synthesis

curl -X POST http://localhost:8000/v1/audio/speech \
  -H "Content-Type: application/json" \
  -d '{"input": "Hello, how are you?"}' \
  --output output.wav

Voice cloning

Supplying the reference transcript (text) materially improves cloning fidelity.

import requests

resp = requests.post(
    "http://localhost:8000/v1/audio/speech",
    json={
        "input": "Have a nice day and enjoy south california sunshine.",
        "references": [{
            "audio_path": "ref.wav",
            "text": "Hey, Adam here. Let's create something that feels real, sounds human, and connects every time.",
        }],
        "temperature": 0.8, "top_k": 50, "max_new_tokens": 1024,
    },
)
with open("output.wav", "wb") as f:
    f.write(resp.content)

Streaming (Server-Sent Events)

Set "stream": true to receive base64-encoded WAV chunks as the vocoder emits them — sub-second time-to-first-audio. Each event carries audio.data (base64 WAV bytes); the terminal event has finish_reason: "stop" plus usage metadata.

import requests, base64, json

with requests.post(
    "http://localhost:8000/v1/audio/speech",
    json={"input": "Get the trust fund to the bank early.", "stream": True},
    stream=True,
) as resp, open("output.wav", "wb") as f:
    for line in resp.iter_lines():
        if not line or not line.startswith(b"data: ") or line == b"data: [DONE]":
            continue
        event = json.loads(line[6:])
        if event.get("finish_reason") == "stop":
            break
        audio = event.get("audio") or {}
        if audio.get("data"):
            f.write(base64.b64decode(audio["data"]))

Inline control tokens

Embed <|emotion:…|>, <|style:…|>, <|prosody:…|>, and <|sfx:…|> tokens directly in input. Two rules:

Delivery tokens first. Emotion, style, and the prosody speed / pitch / expressive tokens shape the whole turn — put them at the start of input. Positional tokens (<|prosody:pause|>, <|prosody:long_pause|>, <|sfx:…|>) go inline exactly where they fire.
Pair every <|sfx:…|> with its onomatopoeia. E.g. <|sfx:laughter|>Haha, <|sfx:sigh|>Uh, <|sfx:sneeze|>Achoo. The written sound gives the model the acoustic cue to realize the effect.

Example — amusement + laughter:

curl -X POST http://localhost:8000/v1/audio/speech \
  -H "Content-Type: application/json" \
  -d '{"input": "<|emotion:amusement|><|prosody:expressive_high|>Wait, wait, that was kind of hilarious. <|sfx:laughter|>Hehe, no, seriously, I was not ready for that."}' \
  --output output.wav

Throughput

Throughput on Seed-TTS EN (full set, N=1088 per run). Client --max-concurrency sweep against a Higgs server (max_running_requests=16, bf16, CUDA Graph on). Each row is the mean of 3 runs. Hardware: 1× H100.

<div style="font-family:-apple-system,BlinkMacSystemFont,'Segoe UI',Roboto,sans-serif;width:100%;padding:16px 0;overflow-x:auto;-webkit-overflow-scrolling:touch;text-align:center"> <table style="display:inline-table;width:max-content;max-width:none;border-collapse:collapse;font-size:clamp(11px,1.4vw,13px);text-align:center"> <thead><tr> <th style="padding:18px 24px;text-align:center;font-weight:600;border-bottom:2px solid #ea580c;color:#ea580c">Concurrency</th> <th style="padding:18px 24px;text-align:center;font-weight:500;white-space:nowrap;border-bottom:2px solid #ea580c;color:#ea580c;font-size:14px">Throughput (req/s)</th> <th style="padding:18px 24px;text-align:center;font-weight:500;white-space:nowrap;border-bottom:2px solid #ea580c;color:#ea580c;font-size:14px">Mean latency</th> <th style="padding:18px 24px;text-align:center;font-weight:500;white-space:nowrap;border-bottom:2px solid #ea580c;color:#ea580c;font-size:14px">RTF (per-req)</th> <th style="padding:18px 24px;text-align:center;font-weight:500;white-space:nowrap;border-bottom:2px solid #ea580c;color:#ea580c;font-size:14px">audio_s/s</th> </tr></thead> <tbody> <tr> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">1</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">1.62</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">617 ms</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">0.147</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">6.89</td> </tr> <tr> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">2</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">2.70</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">742 ms</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">0.180</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">11.37</td> </tr> <tr> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">4</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">5.45</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">733 ms</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">0.177</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">22.84</td> </tr> <tr> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">8</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">8.91</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">898 ms</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">0.217</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">37.38</td> </tr> <tr> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">16</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">14.74</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">1079 ms</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">0.262</td> <td style="padding:16px 24px;text-align:center;border-bottom:1px solid rgba(128,128,128,0.15)">61.84</td> </tr> </tbody> </table> </div>

Concurrency — Maximum number of in-flight client requests (--max-concurrency).
Throughput (req/s) — Completed requests divided by total benchmark wall-clock time.
Mean latency — Average end-to-end time per request (send to full response received).
RTF (per-req) — Average ratio of processing time to generated audio duration per request (<1 is faster than real time).
audio_s/s — Total seconds of audio produced divided by total benchmark wall-clock time.

To reproduce the results, follow the instructions in this script.

vLLM-Omni Usage

You can also serve these weights with vLLM-Omni, which exposes the same OpenAI-compatible /v1/audio/speech API with zero-shot voice cloning.

hf download bosonai/higgs-tts-3-4b

vllm-omni serve bosonai/higgs-tts-3-4b \
  --host 0.0.0.0 --port 8095 \
  --trust-remote-code --omni

curl -X POST http://localhost:8095/v1/audio/speech \
  -H "Content-Type: application/json" \
  -d '{"model": "bosonai/higgs-tts-3-4b", "input": "Hello, how are you?"}' \
  --output output.wav

Plain-text TTS, voice-clone, and benchmark recipes are in the vLLM-Omni Higgs TTS 3 recipe.

API Usage

For zero-ops deployment, use the Boson AI API.

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},
}

Creator Use (Free for Digital Creators)

In addition to research and non-commercial use, the license includes a Creator Use Grant that lets digital creators use Higgs TTS 3 to produce creative content for free — including monetized content.

What's covered

Podcasts, videos, audiobooks, social media posts, and similar creative works
Personal and commercial/monetized creator channels (ad-supported, sponsored, subscription, etc.)

The one requirement: acknowledge Boson AI's Higgs Audio. The acknowledgment must appear in at least one of the following ways:

In the audio — e.g. "This audio was created with Boson AI's Higgs Audio."
In the accompanying text, displayed prominently — e.g. in the post body, video description, or show notes. It must be clearly visible and not hidden at the bottom of the credits or annotations.

Suggested credit string:

This audio was created with Boson AI's Higgs Audio — https://www.boson.ai/higgs-audio

Still requires a separate commercial license. The Creator Use Grant covers creating content with the model. It does not cover hosting the model behind an API or as a service, redistributing/reselling/fine-tuning the model for resale, or embedding the model in a product or application. For these uses, contact us for a commercial license.

The Creator Use Grant does not change the use restrictions — no non-consensual voice cloning or impersonation, no fraud or deception, and AI-generated audio must be disclosed where required. Full terms are in Section II-A of the LICENSE.

License

Boson Higgs TTS 3 Research and Non-Commercial License — see LICENSE. Includes a Creator Use Grant (free monetized creator use with attribution; see Creator Use above).

Have a use case that isn’t covered by the current license? We’d still love to hear from you. Reach out to [email protected] — we’re open to discussing your use case and alternative licensing arrangements.

Task: Text-to-Speech
Library: transformers
Type: higgs_multimodal_qwen3
License: other
Languages: af, ar, as, ast, az, ba
Created: Sep 24, 2026
Updated: Sep 24, 2026