这个榜单回答什么
看自然度、清晰度、情绪和语言覆盖。
不能说明什么
盲听偏好不能替代具体声音、语种和授权要求。
选择模型时
先根据具体生成任务筛选，再回到原始榜单查看样例、投票量和模型版本。
Media Leaderboard
文本转语音榜单
通过音频盲听投票比较语音自然度与质量。
评价口径Quality Elo
当前领先Qwen-Audio-3.0-TTS-Plus
公开结果1,237
| 排名 | 机构 | 模型 | Elo | 95% CI | 样本数 | 竞技场声音 | 发布日期 | API 价格 |
|---|---|---|---|---|---|---|---|---|
| 1范围 1-2 | Alibaba | Qwen-Audio-3.0-TTS-Plus | 1,237 | -16/16 | 1,477 | 8 voices | Jul 2026 | $27.6 /1M chars |
| 2范围 1-2 | SpeechifyAI | Simba 3.2 | 1,230 | -16/16 | 1,470 | 8 voices | Jul 2026 | $10.0 /1M chars |
| 3范围 3-5 | Gemini 3.1 Flash TTS | 1,212 | -13/13 | 2,817 | 7 voices | Apr 2026 | $18.3 /1M chars | |
| 4范围 3-6 | Cartesia | Sonic 3.5 | 1,210 | -14/14 | 2,295 | 8 voices | May 2026 | $49.0 /1M chars |
| 5范围 3-7 | Inworld | Realtime TTS 1.5 Max | 1,200 | -14/14 | 2,049 | 8 voices | Jan 2026 | $26.0 /1M chars |
| 6范围 4-7 | Inworld | Realtime TTS-2 - Research Preview | 1,198 | -14/14 | 2,039 | 8 voices | May 2026 | $20.8 /1M chars |
| 7范围 5-10 | VUI Labs | Luna TTS | 1,186 | -16/16 | 1,351 | 3 voices | Jun 2026 | N/A |
| 8范围 7-11 | MiniMax | Speech 2.8 HD | 1,179 | -12/12 | 3,928 | 8 voices | Feb 2026 | $100.0 /1M chars |
| 9范围 7-11 | StepFun | StepAudio 2.5 TTS | 1,175 | -18/18 | 1,031 | 3 voices | Apr 2026 | $85.0 /1M chars |
| 10范围 8-11 | ElevenLabs | Eleven v3 | 1,173 | -12/12 | 3,783 | 8 voices | Feb 2026 | $100.0 /1M chars |
| 11范围 8-11 | async | Async Flash v1.5 | 1,168 | -15/15 | 1,489 | 8 voices | May 2026 | $10.1 /1M chars |
| 12范围 12-18 | Smallest.ai | Lightning V3.1 Pro TTS (Jun 2026) | 1,150 | -15/15 | 1,391 | 8 voices | Jun 2026 | $19.5 /1M chars |
| 13范围 12-17 | MiniMax | Speech 2.8 Turbo | 1,149 | -12/12 | 3,845 | 8 voices | Feb 2026 | $60.0 /1M chars |
| 14范围 12-19 | async | Async Pro v1.0 | 1,143 | -15/15 | 1,432 | 8 voices | May 2026 | $20.2 /1M chars |
| 15范围 12-19 | StepFun | Step TTS 2 (Mar 2026) | 1,141 | -14/14 | 1,695 | 2 voices | Feb 2026 | $40.0 /1M chars |
| 16范围 12-20 | Fish Audio | Fish Audio S2.1 Pro | 1,139 | -15/15 | 1,432 | 8 voices | Jun 2026 | $15.0 /1M chars |
| 17范围 12-20 | MiniMax | Speech 2.6 HD | 1,138 | -12/12 | 3,582 | 8 voices | Oct 2025 | $100.0 /1M chars |
| 18范围 12-20 | Inworld | Realtime TTS 1.5 Mini | 1,137 | -13/13 | 2,542 | 8 voices | Jan 2026 | $10.4 /1M chars |
| 19范围 15-22 | MiniMax | Speech 2.6 Turbo | 1,130 | -12/12 | 3,933 | 8 voices | Oct 2025 | $60.0 /1M chars |
| 20范围 16-25 | Microsoft | Azure HD 2.5 | 1,126 | -14/14 | 1,941 | 8 voices | Mar 2026 | $22.0 /1M chars |
| 21范围 19-25 | SpeechifyAI | Simba 3.0 | 1,121 | -14/14 | 1,861 | 8 voices | Feb 2026 | $12.5 /1M chars |
| 22范围 15-28 | SpaceXAI | SpaceXAI TTS | 1,120 | -22/22 | 358 | 6 voices | Mar 2026 | $15.0 /1M chars |
| 23范围 19-25 | Fish Audio | Fish Audio S2 Pro开放权重 | 1,117 | -14/14 | 1,580 | 4 voices | Mar 2026 | $15.0 /1M chars |
| 24范围 20-28 | StepFun | Step Audio EditX (Mar 2026)开放权重 | 1,114 | -15/15 | 1,458 | 2 voices | Feb 2026 | N/A |
| 25范围 21-28 | MiniMax | Speech-02-HD | 1,112 | -12/12 | 4,029 | 4 voices | Mar 2025 | $100.0 /1M chars |
| 26范围 25-28 | ElevenLabs | Multilingual v2 | 1,103 | -11/11 | 7,387 | 8 voices | Aug 2023 | $100.0 /1M chars |
| 27范围 24-28 | OpenAI | TTS-1 HD | 1,103 | -12/12 | 3,169 | 6 voices | Nov 2023 | $30.0 /1M chars |
| 28范围 25-28 | ElevenLabs | Turbo v2.5 | 1,103 | -11/11 | 6,937 | 8 voices | Jul 2024 | $50.0 /1M chars |
| 29范围 29-36 | Gradium | Gradium TTS | 1,087 | -15/15 | 1,223 | 8 voices | Jun 2026 | N/A |
| 30范围 29-34 | ElevenLabs | Flash v2.5 | 1,087 | -11/11 | 5,616 | 7 voices | Dec 2024 | $50.0 /1M chars |
| 31范围 29-36 | MiniMax | Speech-02-Turbo | 1,085 | -12/12 | 3,541 | 4 voices | Mar 2025 | $60.0 /1M chars |
| 32范围 29-39 | Resemble AI | Chatterbox HD | 1,083 | -14/14 | 1,981 | — | May 2025 | $40.0 /1M chars |
| 33范围 29-36 | OpenAI | TTS-1 | 1,083 | -11/11 | 6,093 | 6 voices | Nov 2023 | $15.0 /1M chars |
| 34范围 29-43 | Maya Research | Maya 2 Flash | 1,082 | -18/18 | 1,159 | 4 voices | Jul 2026 | N/A |
| 35范围 29-43 | Mistral | Voxtral TTS开放权重 | 1,076 | -14/14 | 1,566 | 3 voices | Mar 2026 | $16.0 /1M chars |
| 36范围 29-43 | Gemini 2.5 Flash Lite TTS | 1,075 | -13/13 | 2,661 | 5 voices | Dec 2025 | $9.2 /1M chars | |
| 37范围 34-45 | Cartesia | Sonic 3 | 1,070 | -12/12 | 3,211 | 8 voices | Oct 2025 | $49.0 /1M chars |
| 38范围 35-46 | Studio | 1,069 | -12/12 | 4,277 | 4 voices | Feb 2023 | $160.0 /1M chars | |
| 39范围 35-46 | Fish Audio | OpenAudio S1 | 1,069 | -12/12 | 4,231 | 8 voices | Jun 2025 | $15.0 /1M chars |
| 40范围 32-49 | OpenAI | GPT-Realtime-2 | 1,068 | -15/15 | 1,262 | 5 voices | May 2026 | $191.6 /1M chars |
| 41范围 35-48 | Amazon | Polly Generative | 1,066 | -12/12 | 3,547 | 4 voices | May 2024 | $30.0 /1M chars |
| 42范围 35-50 | SpeechifyAI | SIMBA 1.6 | 1,064 | -14/14 | 1,840 | 7 voices | Oct 2024 | $10.0 /1M chars |
| 43范围 35-50 | Xiaomi | MiMo-V2.5-TTS | 1,064 | -14/14 | 1,430 | 4 voices | - | N/A |
| 44范围 37-50 | NVIDIA | Magpie-Multilingual 357M (Feb 2026)开放权重 | 1,061 | -14/14 | 1,505 | 5 voices | Feb 2026 | N/A |
| 45范围 37-50 | MiniMax | T2A-01-HD | 1,061 | -12/12 | 4,745 | 8 voices | Jan 2025 | $50.0 /1M chars |
| 46范围 40-50 | Kokoro | Kokoro 82M v1.0开放权重 | 1,058 | -11/11 | 5,220 | 8 voices | Jan 2025 | $0.7 /1M chars |
| 47范围 38-54 | Hume AI | Octave 2 | 1,057 | -13/13 | 2,834 | 8 voices | Oct 2025 | $87.5 /1M chars |
| 48范围 39-57 | Hithink | Speech 2.6 | 1,054 | -15/15 | 1,194 | 8 voices | Jul 2026 | N/A |
| 49范围 37-58 | Maya Research | Maya 2 Global | 1,053 | -18/18 | 1,112 | 4 voices | Jul 2026 | N/A |
| 50范围 44-57 | Chirp 3: HD | 1,052 | -12/12 | 3,123 | 8 voices | Mar 2025 | $30.0 /1M chars | |
| 51范围 46-58 | async | Async Flash v1.0 | 1,046 | -12/12 | 3,699 | 8 voices | Jul 2025 | $10.1 /1M chars |
| 52范围 46-60 | Amazon | Polly Long-Form | 1,046 | -14/14 | 1,626 | 3 voices | Nov 2023 | $100.0 /1M chars |
| 53范围 47-59 | Maya Research | Maya1开放权重 | 1,046 | -12/12 | 3,071 | 4 voices | Jun 2025 | N/A |
| 54范围 42-62 | Fish Audio | OpenAudio S1 Mini | 1,045 | -20/20 | 1,777 | 8 voices | Jun 2025 | $15.0 /1M chars |
| 55范围 47-60 | Journey | 1,043 | -14/14 | 1,864 | 1voice | Dec 2023 | $160.0 /1M chars | |
| 56范围 48-61 | Rime | Coda | 1,041 | -14/14 | 1,533 | 7 voices | May 2026 | $50.0 /1M chars |
| 57范围 50-60 | Cartesia | Sonic English (Oct 2024) | 1,041 | -12/12 | 3,187 | 8 voices | Oct 2024 | $49.0 /1M chars |
| 58范围 51-64 | Gemini 2.5 Flash TTS (Dec 2025) | 1,036 | -13/13 | 2,461 | 8 voices | Dec 2025 | $9.2 /1M chars | |
| 59范围 51-64 | Gemini 2.5 Pro (Dec 2025) | 1,034 | -13/13 | 2,077 | 4 voices | Dec 2025 | $18.3 /1M chars | |
| 60范围 54-64 | SpeechifyAI | SIMBA 1.0 | 1,034 | -12/12 | 4,726 | 7 voices | Jun 2024 | $10.0 /1M chars |
| 61范围 51-67 | Microsoft | Azure Neural | 1,027 | -19/19 | 1,222 | 8 voices | Sept 2018 | $15.0 /1M chars |
| 62范围 58-67 | Microsoft | MAI-Voice-1 | 1,026 | -14/14 | 1,727 | 4 voices | Apr 2026 | N/A |
| 63范围 58-67 | Smallest.ai | Lightning v3.1 | 1,025 | -14/14 | 1,863 | 7 voices | Mar 2026 | $25.0 /1M chars |
| 64范围 59-67 | Hume AI | Octave TTS | 1,024 | -12/12 | 4,661 | 8 voices | Feb 2025 | $87.5 /1M chars |
| 65范围 61-67 | MiniMax | T2A-01-Turbo | 1,021 | -12/12 | 4,893 | 8 voices | Jan 2025 | $30.0 /1M chars |
| 66范围 61-69 | Xiaomi | MiMo-V2-TTS | 1,017 | -15/15 | 1,381 | 1voice | Mar 2026 | N/A |
| 67范围 64-69 | Resemble AI | Chatterbox开放权重 | 1,012 | -12/12 | 4,544 | 8 voices | May 2025 | $25.0 /1M chars |
| 68范围 66-70 | NVIDIA | Magpie-Multilingual 357M开放权重 | 1,005 | -12/12 | 3,430 | 5 voices | Jan 2026 | N/A |
| 69范围 66-70 | Rime | Arcana v3 | 1,004 | -14/14 | 1,783 | 6 voices | Feb 2026 | $50.0 /1M chars |
| 70 | Zyphra | Zonos-v0.1开放权重 | 1,000 | 0/0 | 4,778 | 8 voices | Feb 2025 | $20.0 /1M chars |
| 71范围 71-73 | Murf AI | Murf Speech Gen 2 | 970 | -12/12 | 5,057 | 8 voices | Jul 2024 | $100.0 /1M chars |
| 72范围 71-74 | LMNT | LMNT | 968 | -13/13 | 3,848 | 5 voices | Dec 2023 | $49.0 /1M chars |
| 73范围 71-75 | Microsoft | VibeVoice 1.5B | 967 | -14/14 | 1,899 | 2 voices | Aug 2025 | N/A |
| 74范围 71-75 | OpenVoice | OpenVoice v2开放权重 | 957 | -14/14 | 2,871 | — | Apr 2024 | $8.3 /1M chars |
| 75范围 72-75 | Microsoft | VibeVoice 7B开放权重 | 955 | -14/14 | 1,859 | 3 voices | Aug 2025 | N/A |
| 76范围 76-78 | NVIDIA | Magpie Multilingual | 939 | -15/15 | 1,747 | 3 voices | Apr 2025 | N/A |
| 77范围 76-78 | Neuphonic | Neuphonic TTS | 933 | -14/14 | 2,083 | 7 voices | Oct 2024 | $20.8 /1M chars |
| 78范围 76-79 | Alibaba | Qwen3 TTS Flash | 930 | -15/15 | 1,848 | 4 voices | Sept 2025 | $10.0 /1M chars |
| 79范围 78-80 | Alibaba | Qwen3 TTS | 918 | -14/14 | 2,492 | 8 voices | - | N/A |
| 80范围 79-81 | Coqui | XTTS v2开放权重 | 915 | -15/15 | 2,357 | — | Nov 2023 | $40.4 /1M chars |
| 81 | WaveNet | 902 | -12/12 | 6,408 | 8 voices | Sept 2016 | $16.0 /1M chars | |
| 82范围 81-86 | StyleTTS | StyleTTS 2开放权重 | 887 | -16/16 | 2,288 | — | Jun 2023 | $2.8 /1M chars |
| 83范围 82-86 | Rime | Mist V2 | 884 | -15/15 | 2,651 | 7 voices | Mar 2025 | $31.1 /1M chars |
| 84范围 82-86 | Neural2 | 883 | -12/12 | 6,647 | 8 voices | Jun 2022 | $16.0 /1M chars | |
| 85范围 82-86 | Amazon | Polly Neural | 879 | -15/15 | 2,631 | 7 voices | Jul 2019 | $16.0 /1M chars |
| 86范围 82-86 | Standard | 877 | -12/12 | 6,768 | 8 voices | Aug 2018 | $4.0 /1M chars | |
| 87范围 87-88 | Murf AI | Falcon (Beta) | 854 | -15/15 | 2,743 | 7 voices | Nov 2025 | N/A |
| 88范围 87-88 | Noiz | Noiz TTS | 851 | -17/17 | 1,374 | 8 voices | Nov 2025 | N/A |
| 89 | MetaVoice | MetaVoice v1开放权重 | 833 | -18/18 | 1,259 | — | Feb 2024 | N/A |
| 90 | Amazon | Polly Standard | 811 | -16/16 | 3,461 | 4 voices | Nov 2016 | $4.0 /1M chars |