提供商语音 Arena 排行榜Artificial Analysis
使用各提供商自己的原生语音比较文本转语音(TTS)模型。
Range | Creator | Model | Elo | 95% CI | Samples | Arena 语音 | 发布 | API 定价 1 | |
|---|---|---|---|---|---|---|---|---|---|
| 1 | 1-2 | Sonic 3.6 | 1272 | -17/17 | 1,740 | Aug 2026 | US$49.0 /100 万字符 | ||
| 2 | 1-3 | Qwen-Audio-3.0-TTS-Plus | 1260 | -17/17 | 1,438 | Sept 2026 | US$27.6 /100 万字符 | ||
| 3 | 2-5 | Realtime TTS-2 | 1245 | -18/18 | 1,229 | Aug 2026 | US$20.8 /100 万字符 | ||
| 4 | 3-5 | Simba 3.2 | 1237 | -14/14 | 2,381 | Jul 2026 | US$6.6 /100 万字符 | ||
| 5 | 4-5 | Luna TTS | 1230 | -14/14 | 2,502 | Jun 2026 | US$15.0 /100 万字符 | ||
| 6 | 6-10 | Realtime TTS-2 Flash | 1210 | -15/15 | 1,677 | Aug 2026 | US$10.4 /100 万字符 | ||
| 7 | 6-10 | Breeze TTS 2 | 1204 | -16/16 | 1,374 | Aug 2026 | US$34.0 /100 万字符 | ||
| 8 | 6-11 | StepAudio 2.5 TTS (Aug 2026) | 1199 | -17/17 | 1,262 | Aug 2026 | US$85.0 /100 万字符 | ||
| 9 | 6-10 | Gemini 3.1 Flash TTS | 1199 | -12/12 | 3,415 | Apr 2026 | US$18.3 /100 万字符 | ||
| 10 | 6-11 | v3 Conversational | 1197 | -15/15 | 1,829 | Aug 2026 | US$50.0 /100 万字符 | ||
| 11 | 10-13 | Sonic 3.5 | 1186 | -12/12 | 2,982 | May 2026 | US$49.0 /100 万字符 | ||
| 12 | 11-15 | Lightning V3.1 Pro (Jul 2026) | 1175 | -14/14 | 1,902 | Jul 2026 | US$19.5 /100 万字符 | ||
| 13 | 11-15 | Soniox TTS Real-Time v2 | 1175 | -12/12 | 4,601 | Aug 2026 | US$14.2 /100 万字符 | ||
| 14 | 12-15 | Eleven v3 | 1167 | -11/11 | 4,328 | Feb 2026 | US$100.0 /100 万字符 | ||
| 15 | 12-16 | Speech 2.8 HD | 1166 | -11/11 | 4,402 | Feb 2026 | US$100.0 /100 万字符 | ||
| 16 | 14-18 | Falcon 2 | 1155 | -15/15 | 1,722 | Jul 2026 | US$10.0 /100 万字符 | ||
| 17 | 16-19 | Gradium TTS (Aug 2026) | 1150 | -14/14 | 1,650 | Aug 2026 | US$47.2 /100 万字符 | ||
| 18 | 16-18 | Speech 2.8 Turbo | 1149 | -11/11 | 4,311 | Feb 2026 | US$60.0 /100 万字符 | ||
| 19 | 17-24 | Fish Audio S2.1 Pro | 1138 | -13/13 | 2,147 | Jun 2026 | US$15.0 /100 万字符 | ||
| 20 | 19-25 | Step TTS 2 (Mar 2026) | 1133 | -13/13 | 2,203 | Feb 2026 | US$40.0 /100 万字符 | ||
| 21 | 19-26 | Async Flash v1.5 | 1130 | -13/13 | 2,083 | May 2026 | US$10.1 /100 万字符 | ||
| 22 | 19-26 | Azure HD 2.5 | 1129 | -12/12 | 2,396 | Mar 2026 | US$22.0 /100 万字符 | ||
| 23 | 19-26 | Speech 2.6 HD | 1127 | -12/12 | 3,824 | Oct 2025 | US$100.0 /100 万字符 | ||
| 24 | 19-29 | SpaceXAI TTS | 1126 | -15/15 | 1,204 | Mar 2026 | US$15.0 /100 万字符 | ||
| 25 | 20-29 | Fish Audio S2 Pro | 1121 | -13/13 | 2,205 | Mar 2026 | US$15.0 /100 万字符 | ||
| 26 | 21-29 | Simba 3.0 | 1118 | -12/12 | 2,471 | Feb 2026 | US$6.6 /100 万字符 | ||
| 27 | 25-29 | Speech 2.6 Turbo | 1114 | -11/11 | 4,166 | Oct 2025 | US$60.0 /100 万字符 | ||
| 28 | 24-29 | Async Pro v1.0 | 1113 | -13/13 | 1,993 | May 2026 | US$20.2 /100 万字符 | ||
| 29 | 25-30 | Speech-02-HD | 1112 | -12/12 | 4,239 | Mar 2025 | US$100.0 /100 万字符 | ||
| 30 | 29-35 | TTS-1 HD | 1100 | -12/12 | 3,460 | Nov 2023 | US$30.0 /100 万字符 | ||
| 31 | 30-37 | Chatterbox HD | 1096 | -13/13 | 2,295 | — | May 2025 | US$40.0 /100 万字符 | |
| 32 | 30-37 | Speech 2.6 | 1096 | -13/13 | 1,937 | Jul 2026 | 暂无 | ||
| 33 | 30-37 | Step Audio EditX (Mar 2026) | 1094 | -13/13 | 1,923 | Feb 2026 | 暂无 | ||
| 34 | 30-37 | Turbo v2.5 | 1094 | -11/11 | 6,745 | Jul 2024 | US$50.0 /100 万字符 | ||
| 35 | 30-37 | Multilingual v2 | 1092 | -10/10 | 7,197 | Aug 2023 | US$100.0 /100 万字符 | ||
| 36 | 31-41 | Speech-02-Turbo | 1086 | -12/12 | 3,697 | Mar 2025 | US$60.0 /100 万字符 | ||
| 37 | 32-41 | TTS-1 | 1085 | -11/11 | 6,024 | Nov 2023 | US$15.0 /100 万字符 | ||
| 38 | 36-44 | Gemini 2.5 Flash Lite TTS | 1080 | -12/12 | 3,053 | Dec 2025 | US$9.2 /100 万字符 | ||
| 39 | 36-45 | OpenAudio S1 | 1077 | -11/11 | 4,411 | Jun 2025 | US$15.0 /100 万字符 | ||
| 40 | 36-46 | Voxtral TTS | 1075 | -13/13 | 2,028 | Mar 2026 | US$16.0 /100 万字符 | ||
| 41 | 36-45 | Studio | 1075 | -12/12 | 4,349 | Feb 2023 | US$160.0 /100 万字符 | ||
| 42 | 38-45 | Flash v2.5 | 1074 | -11/11 | 5,597 | Dec 2024 | US$50.0 /100 万字符 | ||
| 43 | 37-49 | Journey | 1072 | -13/13 | 2,167 | Dec 2023 | US$160.0 /100 万字符 | ||
| 44 | 38-49 | GPT-Realtime-2 | 1072 | -13/13 | 1,967 | May 2026 | US$191.6 /100 万字符 | ||
| 45 | 38-51 | Maya 2 Flash | 1068 | -14/14 | 1,807 | Jul 2026 | 暂无 | ||
| 46 | 40-53 | Magpie-Multilingual 357M (Feb 2026) | 1063 | -13/13 | 1,959 | Feb 2026 | 暂无 | ||
| 47 | 43-53 | Polly Generative | 1062 | -12/12 | 3,640 | May 2024 | US$30.0 /100 万字符 | ||
| 48 | 43-51 | T2A-01-HD | 1062 | -11/11 | 4,865 | Jan 2025 | US$50.0 /100 万字符 | ||
| 49 | 44-53 | Kokoro 82M v1.0 | 1061 | -11/11 | 5,224 | Jan 2025 | US$0.7 /100 万字符 | ||
| 50 | 45-57 | Sonic 3 | 1057 | -12/12 | 3,407 | Oct 2025 | US$49.0 /100 万字符 | ||
| 51 | 46-57 | Coda | 1054 | -13/13 | 2,179 | May 2026 | US$50.0 /100 万字符 | ||
| 52 | 46-59 | Gemini 2.5 Flash TTS (Dec 2025) | 1051 | -12/12 | 2,796 | Dec 2025 | US$9.2 /100 万字符 | ||
| 53 | 46-59 | Chirp 3: HD | 1051 | -12/12 | 3,260 | Mar 2025 | US$30.0 /100 万字符 | ||
| 54 | 49-59 | Async Flash v1.0 | 1049 | -12/12 | 3,937 | Jul 2025 | US$10.1 /100 万字符 | ||
| 55 | 47-59 | MiMo-V2.5-TTS | 1049 | -13/13 | 1,855 | - | 暂无 | ||
| 56 | 50-59 | Octave 2 | 1048 | -12/12 | 3,011 | Oct 2025 | US$87.5 /100 万字符 | ||
| 57 | 50-61 | Maya 2 Global | 1047 | -14/14 | 1,662 | Jul 2026 | 暂无 | ||
| 58 | 50-68 | OpenAudio S1 Mini | 1041 | -20/20 | 1,681 | Jun 2025 | US$15.0 /100 万字符 | ||
| 59 | 52-64 | Maya1 | 1040 | -12/12 | 3,191 | Jun 2025 | 暂无 | ||
| 60 | 57-67 | Polly Long-Form | 1034 | -13/13 | 1,880 | Nov 2023 | US$100.0 /100 万字符 | ||
| 61 | 58-69 | Bland Speech v3 | 1033 | -13/13 | 2,233 | Aug 2026 | US$40.0 /100 万字符 | ||
| 62 | 58-70 | Higgs Audio V3 TTS | 1032 | -14/14 | 1,957 | Jun 2026 | US$15.0 /100 万字符 | ||
| 63 | 54-70 | Azure Neural | 1031 | -19/19 | 1,158 | Sept 2018 | US$15.0 /100 万字符 | ||
| 64 | 58-70 | MAI-Voice-1 | 1030 | -13/13 | 2,126 | Apr 2026 | 暂无 | ||
| 65 | 60-70 | Octave TTS | 1027 | -12/12 | 4,606 | Feb 2025 | US$87.5 /100 万字符 | ||
| 66 | 59-70 | Gemini 2.5 Pro (Dec 2025) | 1027 | -13/13 | 2,401 | Dec 2025 | US$18.3 /100 万字符 | ||
| 67 | 60-70 | T2A-01-Turbo | 1027 | -11/11 | 4,893 | Jan 2025 | US$30.0 /100 万字符 | ||
| 68 | 60-70 | MiMo-V2-TTS | 1021 | -14/14 | 1,709 | Mar 2026 | 暂无 | ||
| 69 | 62-70 | Chatterbox | 1021 | -12/12 | 4,595 | May 2025 | US$25.0 /100 万字符 | ||
| 70 | 62-70 | Lightning v3.1 | 1020 | -13/13 | 2,086 | Mar 2026 | US$25.0 /100 万字符 | ||
| 71 | 71-73 | Raon SpeechLM | 1002 | -14/14 | 2,025 | Apr 2026 | 暂无 | ||
| 72 | 71-73 | Arcana v3 | 1001 | -13/13 | 2,171 | Feb 2026 | US$50.0 /100 万字符 | ||
| 73 | 73 | Zonos-v0.1 | 1000 | 0/0 | 4,792 | Feb 2025 | US$20.0 /100 万字符 | ||
| 74 | 74 | Murf Speech Gen 2 | 982 | -12/12 | 4,958 | Jul 2024 | US$100.0 /100 万字符 | ||
| 75 | 75-77 | VibeVoice 1.5B | 951 | -14/14 | 2,087 | Aug 2025 | 暂无 | ||
| 76 | 75-78 | OpenVoice v2 | 946 | -14/14 | 2,959 | Apr 2024 | US$8.3 /100 万字符 | ||
| 77 | 75-79 | Qwen3 TTS Flash | 940 | -14/14 | 2,107 | Sept 2025 | US$10.0 /100 万字符 | ||
| 78 | 76-80 | Neuphonic TTS | 936 | -14/14 | 2,376 | Oct 2024 | US$20.8 /100 万字符 | ||
| 79 | 77-80 | Kugel 3 | 931 | -13/13 | 4,128 | Jun 2026 | 暂无 | ||
| 80 | 78-82 | Qwen3 TTS | 926 | -13/13 | 2,854 | Jan 2026 | US$20.0 /100 万字符 | ||
| 81 | 80-83 | WaveNet | 915 | -12/12 | 6,214 | Sept 2016 | US$4.0 /100 万字符 | ||
| 82 | 80-83 | XTTS v2 | 914 | -15/15 | 2,514 | — | Nov 2023 | US$40.4 /100 万字符 | |
| 83 | 81-85 | Mist V2 | 906 | -15/15 | 2,813 | Mar 2025 | US$31.1 /100 万字符 | ||
| 84 | 84-88 | Neural2 | 892 | -12/12 | 6,449 | Jun 2022 | US$16.0 /100 万字符 | ||
| 85 | 83-88 | StyleTTS 2 | 891 | -15/15 | 2,424 | — | Jun 2023 | US$2.8 /100 万字符 | |
| 86 | 84-88 | Polly Neural | 889 | -14/14 | 2,728 | Jul 2019 | US$16.0 /100 万字符 | ||
| 87 | 84-88 | Standard | 882 | -12/12 | 6,487 | Aug 2018 | US$4.0 /100 万字符 | ||
| 88 | 84-88 | Noiz TTS | 881 | -15/15 | 1,952 | Nov 2025 | 暂无 | ||
| 89 | 89-90 | MetaVoice v1 | 843 | -18/18 | 1,343 | — | Feb 2024 | 暂无 | |
| 90 | 89-90 | Polly Standard | 827 | -16/16 | 3,277 | Nov 2016 | US$4.0 /100 万字符 |
1 API 定价反映在模型创建者的 API 上使用模型默认设置生成 100 万字符的成本
2 开放权重包括可使用可下载的模型权重或提供商提供的推理运行时进行自托管的模型。可商用开放权重包括已发布许可证允许商业使用(包括需遵守许可条件的使用,例如 NVIDIA AI Enterprise 订阅)的模型,不包括需要另行协商商业协议的模型。
常见问题
按 Elo 评分排名的顶尖文本转语音模型是:1. Sonic 3.6(Elo 1272)、2. Qwen-Audio-3.0-TTS-Plus(Elo 1260)、3. Realtime TTS-2(Elo 1245)、4. Simba 3.2(Elo 1237)和5. Luna TTS(Elo 1230)。排名基于 Speech Arena 中的盲测用户投票。
Kokoro 82M v1.0 最实惠,价格为每 100 万字符 $0.65,Elo 分数为 1061。其他实惠选项包括 StyleTTS 2,价格为每 100 万字符 $2.82。
排名前 5 的开放权重文本转语音模型是:1. Breeze TTS 2(Elo 1204)、2. Fish Audio S2 Pro(Elo 1121)、3. Step Audio EditX (Mar 2026)(Elo 1094)、4. Voxtral TTS(Elo 1075)和5. Magpie-Multilingual 357M (Feb 2026)(Elo 1063)。