Ranking da Arena de voz do provedorArtificial Analysis
Compare modelos de texto para fala (TTS) usando as vozes nativas de cada provedor.
Range | Creator | Model | Elo | 95% CI | Samples | Vozes da Arena | Lançamento | Preço da API 1 | |
|---|---|---|---|---|---|---|---|---|---|
| 1 | 1 |  Cartesia | Sonic 3.6 | 1276 | -16/16 | 1,822 | Aug 2026 | US$ 49,0 /1 mi de caracteres | |
| 2 | 2-3 |  Alibaba | Qwen-Audio-3.0-TTS-Plus | 1260 | -16/16 | 1,559 | Sept 2026 | US$ 27,6 /1 mi de caracteres | |
| 3 | 2-5 |  Inworld | Realtime TTS-2 | 1247 | -18/18 | 1,262 | Aug 2026 | US$ 20,8 /1 mi de caracteres | |
| 4 | 3-5 |  SpeechifyAI | Simba 3.2 | 1240 | -14/14 | 2,498 | Jul 2026 | US$ 6,6 /1 mi de caracteres | |
| 5 | 4-5 |  VUI Labs | Luna TTS | 1231 | -13/13 | 2,698 | Jun 2026 | US$ 80,0 /1 mi de caracteres | |
| 6 | 6-9 | Inworld | Realtime TTS-2 Flash | 1215 | -15/15 | 1,834 | Aug 2026 | US$ 10,4 /1 mi de caracteres | |
| 7 | 6-10 |  StepFun | StepAudio 2.5 TTS (Aug 2026) | 1209 | -16/16 | 1,326 | Aug 2026 | US$ 85,0 /1 mi de caracteres | |
| 8 | 6-10 |  BreezeBlue | Breeze TTS 2  Pesos abertos | 1205 | -16/16 | 1,453 | Aug 2026 | US$ 34,0 /1 mi de caracteres | |
| 9 | 7-10 |  Google | Gemini 3.1 Flash TTS | 1201 | -12/12 | 3,620 | Apr 2026 | US$ 18,3 /1 mi de caracteres | |
| 10 | 7-11 |  ElevenLabs | v3 Conversational | 1197 | -15/15 | 1,935 | Aug 2026 | US$ 50,0 /1 mi de caracteres | |
| 11 | 11-13 | Cartesia | Sonic 3.5 | 1184 | -12/12 | 3,252 | May 2026 | US$ 49,0 /1 mi de caracteres | |
| 12 | 11-15 |  Smallest.ai | Lightning V3.1 Pro (Jul 2026) | 1179 | -14/14 | 2,118 | Jul 2026 | US$ 19,5 /1 mi de caracteres | |
| 13 | 11-15 |  Soniox | Soniox TTS Real-Time v2 | 1176 | -11/11 | 5,280 | Aug 2026 | US$ 14,2 /1 mi de caracteres | |
| 14 | 12-16 | ElevenLabs | Eleven v3 | 1168 | -11/11 | 4,577 | Feb 2026 | US$ 100,0 /1 mi de caracteres | |
| 15 | 13-16 |  MiniMax | Speech 2.8 HD | 1166 | -11/11 | 4,632 | Feb 2026 | US$ 100,0 /1 mi de caracteres | |
| 16 | 14-18 |  Murf AI | Falcon 2 | 1158 | -14/14 | 1,834 | Jul 2026 | US$ 10,0 /1 mi de caracteres | |
| 17 | 16-18 | MiniMax | Speech 2.8 Turbo | 1151 | -11/11 | 4,513 | Feb 2026 | US$ 60,0 /1 mi de caracteres | |
| 18 | 16-20 |  Gradium | Gradium TTS (Aug 2026) | 1148 | -14/14 | 1,771 | Aug 2026 | US$ 47,2 /1 mi de caracteres | |
| 19 | 17-24 |  Fish Audio | Fish Audio S2.1 Pro | 1139 | -13/13 | 2,359 | Jun 2026 | US$ 15,0 /1 mi de caracteres | |
| 20 | 18-24 | StepFun | Step TTS 2 (Mar 2026) | 1138 | -12/12 | 2,449 | Feb 2026 | US$ 40,0 /1 mi de caracteres | |
| 21 | 19-25 |  Microsoft | Azure HD 2.5 | 1133 | -12/12 | 2,636 | Mar 2026 | US$ 22,0 /1 mi de caracteres | |
| 22 | 19-27 |  SpaceXAI | SpaceXAI TTS | 1131 | -15/15 | 1,355 | Mar 2026 | US$ 15,0 /1 mi de caracteres | |
| 23 | 20-27 | MiniMax | Speech 2.6 HD | 1127 | -11/11 | 4,050 | Oct 2025 | US$ 100,0 /1 mi de caracteres | |
| 24 | 19-27 |  async | Async Flash v1.5 | 1127 | -13/13 | 2,346 | May 2026 | US$ 10,1 /1 mi de caracteres | |
| 25 | 21-29 | Fish Audio | Fish Audio S2 Pro Pesos abertos | 1122 | -12/12 | 2,406 | Mar 2026 | US$ 15,0 /1 mi de caracteres | |
| 26 | 23-29 | SpeechifyAI | Simba 3.0 | 1119 | -12/12 | 2,700 | Feb 2026 | US$ 6,6 /1 mi de caracteres | |
| 27 | 23-29 | MiniMax | Speech 2.6 Turbo | 1118 | -11/11 | 4,427 | Oct 2025 | US$ 60,0 /1 mi de caracteres | |
| 28 | 25-30 | async | Async Pro v1.0 | 1113 | -13/13 | 2,229 | May 2026 | US$ 20,2 /1 mi de caracteres | |
| 29 | 25-30 | MiniMax | Speech-02-HD | 1111 | -11/11 | 4,411 | Mar 2025 | US$ 100,0 /1 mi de caracteres | |
| 30 | 28-33 |  OpenAI | TTS-1 HD | 1105 | -11/11 | 3,648 | Nov 2023 | US$ 30,0 /1 mi de caracteres | |
| 31 | 30-36 | StepFun | Step Audio EditX (Mar 2026) Pesos abertos | 1098 | -13/13 | 2,113 | Feb 2026 | N/D | |
| 32 | 30-36 |  Hithink | Speech 2.6 | 1097 | -13/13 | 2,029 | Jul 2026 | N/D | |
| 33 | 31-36 | ElevenLabs | Turbo v2.5 | 1094 | -10/10 | 6,938 | Jul 2024 | US$ 50,0 /1 mi de caracteres | |
| 34 | 30-40 |  Resemble AI | Chatterbox HD | 1093 | -13/13 | 2,410 | — | May 2025 | US$ 40,0 /1 mi de caracteres |
| 35 | 31-38 | ElevenLabs | Multilingual v2 | 1092 | -10/10 | 7,374 | Aug 2023 | US$ 100,0 /1 mi de caracteres | |
| 36 | 33-43 | OpenAI | TTS-1 | 1086 | -11/11 | 6,162 | Nov 2023 | US$ 15,0 /1 mi de caracteres | |
| 37 | 33-43 | MiniMax | Speech-02-Turbo | 1084 | -12/12 | 3,832 | Mar 2025 | US$ 60,0 /1 mi de caracteres | |
| 38 | 33-43 | Google | Gemini 2.5 Flash Lite TTS | 1084 | -12/12 | 3,206 | Dec 2025 | US$ 9,2 /1 mi de caracteres | |
| 39 | 36-43 | Google | Studio | 1081 | -11/11 | 4,508 | Feb 2023 | US$ 160,0 /1 mi de caracteres | |
| 40 | 36-43 | Fish Audio | OpenAudio S1 | 1080 | -11/11 | 4,526 | Jun 2025 | US$ 15,0 /1 mi de caracteres | |
| 41 | 35-45 | Google | Journey | 1080 | -13/13 | 2,279 | Dec 2023 | US$ 160,0 /1 mi de caracteres | |
| 42 | 36-45 | ElevenLabs | Flash v2.5 | 1078 | -11/11 | 5,737 | Dec 2024 | US$ 50,0 /1 mi de caracteres | |
| 43 | 36-47 |  Mistral | Voxtral TTS Pesos abertos | 1075 | -13/13 | 2,184 | Mar 2026 | US$ 16,0 /1 mi de caracteres | |
| 44 | 40-51 | OpenAI | GPT-Realtime-2 | 1068 | -13/13 | 2,109 | May 2026 | US$ 191,6 /1 mi de caracteres | |
| 45 | 40-52 |  Maya Research | Maya 2 Flash | 1067 | -14/14 | 1,907 | Jul 2026 | N/D | |
| 46 | 43-52 | MiniMax | T2A-01-HD | 1065 | -11/11 | 4,972 | Jan 2025 | US$ 50,0 /1 mi de caracteres | |
| 47 | 43-54 |  Amazon | Polly Generative | 1064 | -12/12 | 3,733 | May 2024 | US$ 30,0 /1 mi de caracteres | |
| 48 | 44-55 |  Kokoro | Kokoro 82M v1.0 Pesos abertos | 1062 | -11/11 | 5,326 | Jan 2025 | US$ 0,7 /1 mi de caracteres | |
| 49 | 44-56 |  NVIDIA | Magpie-Multilingual 357M (Feb 2026) Pesos abertos | 1061 | -13/13 | 2,083 | Feb 2026 | N/D | |
| 50 | 44-56 |  Rime | Coda | 1061 | -13/13 | 2,294 | May 2026 | US$ 50,0 /1 mi de caracteres | |
| 51 | 44-57 | Cartesia | Sonic 3 | 1059 | -12/12 | 3,570 | Oct 2025 | US$ 49,0 /1 mi de caracteres | |
| 52 | 46-58 | Google | Chirp 3: HD | 1054 | -12/12 | 3,420 | Mar 2025 | US$ 30,0 /1 mi de caracteres | |
| 53 | 46-59 |  Hume AI | Octave 2 | 1053 | -12/12 | 3,105 | Oct 2025 | US$ 87,5 /1 mi de caracteres | |
| 54 | 47-59 | Google | Gemini 2.5 Flash TTS (Dec 2025) | 1053 | -12/12 | 2,922 | Dec 2025 | US$ 9,2 /1 mi de caracteres | |
| 55 | 47-60 |  Xiaomi | MiMo-V2.5-TTS | 1052 | -13/13 | 2,024 | - | N/D | |
| 56 | 48-62 | Maya Research | Maya 2 Global | 1050 | -14/14 | 1,780 | Jul 2026 | N/D | |
| 57 | 52-62 | async | Async Flash v1.0 | 1048 | -11/11 | 4,085 | Jul 2025 | US$ 10,1 /1 mi de caracteres | |
| 58 | 52-65 | Amazon | Polly Long-Form | 1043 | -13/13 | 1,960 | Nov 2023 | US$ 100,0 /1 mi de caracteres | |
| 59 | 48-67 | Fish Audio | OpenAudio S1 Mini Pesos abertos | 1042 | -20/20 | 1,681 | Jun 2025 | US$ 15,0 /1 mi de caracteres | |
| 60 | 56-67 | Maya Research | Maya1 Pesos abertos | 1040 | -12/12 | 3,285 | Jun 2025 | N/D | |
| 61 | 56-67 |  Bland AI | Bland Speech v3 | 1038 | -13/13 | 2,348 | Aug 2026 | US$ 40,0 /1 mi de caracteres | |
| 62 | 56-67 |  Boson AI | Higgs Audio V3 TTS Pesos abertos | 1038 | -13/13 | 2,076 | Jun 2026 | US$ 15,0 /1 mi de caracteres | |
| 63 | 56-70 | Microsoft | Azure Neural | 1033 | -19/19 | 1,158 | Sept 2018 | US$ 15,0 /1 mi de caracteres | |
| 64 | 58-70 | Microsoft | MAI-Voice-1 | 1031 | -13/13 | 2,275 | Apr 2026 | N/D | |
| 65 | 58-70 | Google | Gemini 2.5 Pro (Dec 2025) | 1031 | -12/12 | 2,520 | Dec 2025 | US$ 18,3 /1 mi de caracteres | |
| 66 | 60-70 | MiniMax | T2A-01-Turbo | 1029 | -11/11 | 4,991 | Jan 2025 | US$ 30,0 /1 mi de caracteres | |
| 67 | 60-70 | Hume AI | Octave TTS | 1029 | -11/11 | 4,697 | Feb 2025 | US$ 87,5 /1 mi de caracteres | |
| 68 | 63-70 | Xiaomi | MiMo-V2-TTS | 1021 | -14/14 | 1,805 | Mar 2026 | N/D | |
| 69 | 63-70 | Smallest.ai | Lightning v3.1 | 1021 | -13/13 | 2,215 | Mar 2026 | US$ 25,0 /1 mi de caracteres | |
| 70 | 64-70 | Resemble AI | Chatterbox  Pesos abertos | 1021 | -11/11 | 4,674 | May 2025 | US$ 25,0 /1 mi de caracteres | |
| 71 | 70-73 |  Krafton | Raon SpeechLM | 1007 | -14/14 | 2,047 | Apr 2026 | N/D | |
| 72 | 71-73 | Rime | Arcana v3 | 1006 | -13/13 | 2,272 | Feb 2026 | US$ 50,0 /1 mi de caracteres | |
| 73 | 73 |  Zyphra | Zonos-v0.1 Pesos abertos | 1000 | 0/0 | 4,880 | Feb 2025 | US$ 20,0 /1 mi de caracteres | |
| 74 | 74 | Murf AI | Murf Speech Gen 2 | 979 | -12/12 | 5,045 | Jul 2024 | US$ 100,0 /1 mi de caracteres | |
| 75 | 75-76 | Microsoft | VibeVoice 1.5B Pesos abertos | 961 | -14/14 | 2,148 | Aug 2025 | N/D | |
| 76 | 75-78 |  OpenVoice | OpenVoice v2 Pesos abertos | 951 | -13/13 | 2,994 | Apr 2024 | US$ 8,3 /1 mi de caracteres | |
| 77 | 76-79 | Alibaba | Qwen3 TTS Flash | 945 | -14/14 | 2,167 | Sept 2025 | US$ 10,0 /1 mi de caracteres | |
| 78 | 76-80 |  Neuphonic | Neuphonic TTS | 939 | -14/14 | 2,429 | Oct 2024 | US$ 20,8 /1 mi de caracteres | |
| 79 | 77-80 |  KugelAudio | Kugel 3 | 935 | -13/13 | 4,146 | Jun 2026 | N/D | |
| 80 | 78-82 | Alibaba | Qwen3 TTS | 927 | -13/13 | 2,943 | Jan 2026 | US$ 20,0 /1 mi de caracteres | |
| 81 | 80-82 | Google | WaveNet | 918 | -12/12 | 6,266 | Sept 2016 | US$ 4,0 /1 mi de caracteres | |
| 82 | 80-83 |  Coqui | XTTS v2 Pesos abertos | 916 | -15/15 | 2,557 | — | Nov 2023 | US$ 40,4 /1 mi de caracteres |
| 83 | 81-86 | Rime | Mist V2 | 905 | -15/15 | 2,877 | Mar 2025 | US$ 31,1 /1 mi de caracteres | |
| 84 | 83-88 |  StyleTTS | StyleTTS 2 Pesos abertos | 894 | -15/15 | 2,461 | — | Jun 2023 | US$ 2,8 /1 mi de caracteres |
| 85 | 83-87 | Google | Neural2 | 893 | -12/12 | 6,520 | Jun 2022 | US$ 16,0 /1 mi de caracteres | |
| 86 | 84-88 | Amazon | Polly Neural | 891 | -14/14 | 2,794 | Jul 2019 | US$ 16,0 /1 mi de caracteres | |
| 87 | 84-88 | Google | Standard | 884 | -12/12 | 6,529 | Aug 2018 | US$ 4,0 /1 mi de caracteres | |
| 88 | 84-88 |  Noiz | Noiz TTS | 880 | -15/15 | 2,035 | Nov 2025 | N/D | |
| 89 | 89-90 |  MetaVoice | MetaVoice v1 Pesos abertos | 836 | -18/18 | 1,371 | — | Feb 2024 | N/D |
| 90 | 89-90 | Amazon | Polly Standard | 829 | -16/16 | 3,359 | Nov 2016 | US$ 4,0 /1 mi de caracteres |
1 O preço da API reflete o custo de gerar 1M de caracteres na API do criador do modelo com as configurações padrão do modelo
2 Pesos abertos inclui modelos que podem ser auto-hospedados usando pesos de modelo para download ou um ambiente de inferência fornecido pelo provedor. Pesos abertos para uso comercial inclui modelos cuja licença publicada permite o uso comercial, inclusive sujeito a condições da licença (por exemplo, uma assinatura do NVIDIA AI Enterprise), e exclui modelos que exigem um acordo comercial negociado separadamente.
Perguntas frequentes
Os melhores modelos de texto para fala por classificação Elo são: 1. Sonic 3.6 (Elo 1276), 2. Qwen-Audio-3.0-TTS-Plus (Elo 1260), 3. Realtime TTS-2 (Elo 1247), 4. Simba 3.2 (Elo 1240) e 5. Luna TTS (Elo 1231). Os rankings se baseiam em votos cegos de usuários na Speech Arena.
Kokoro 82M v1.0 é o mais acessível a $0.65 por 1M de caracteres, com uma pontuação Elo de 1062. Outras opções acessíveis incluem StyleTTS 2 a $2.82 por 1M de caracteres.
Os 5 melhores modelos de texto para fala de pesos abertos são: 1. Breeze TTS 2 (Elo 1205), 2. Fish Audio S2 Pro (Elo 1122), 3. Step Audio EditX (Mar 2026) (Elo 1098), 4. Voxtral TTS (Elo 1075) e 5. Kokoro 82M v1.0 (Elo 1062).