Ranking da Arena de voz controladaArtificial Analysis

Compare modelos de texto para fala (TTS) usando as mesmas 8 vozes clonadas (4 dos EUA, 4 do Reino Unido).

Range
Creator
Model
Elo
95% CI
Samples
Lançamento
Preço da API 1
11-2
Inworld logoInworld
Realtime TTS-2
1144-15/152,224Aug 2026US$ 20,8 /1 mi de caracteres
21-2
Cartesia logoCartesia
Sonic 3.6
1139-15/152,233Aug 2026US$ 49,0 /1 mi de caracteres
32-4
Alibaba logoAlibaba
Qwen-Audio-3.0-TTS-Plus
1124-16/162,009Sept 2026US$ 27,6 /1 mi de caracteres
43-4
Cartesia logoCartesia
Sonic 3.5
1119-13/134,138May 2026US$ 49,0 /1 mi de caracteres
55-7
ElevenLabs logoElevenLabs
Eleven v3
1074-13/133,581Feb 2026US$ 100,0 /1 mi de caracteres
65-7
Inworld logoInworld
Realtime TTS-2 Flash
1071-14/142,370Aug 2026US$ 10,4 /1 mi de caracteres
75-7
StepFun logoStepFun
StepAudio 2.5 TTS
1070-14/142,919Apr 2026US$ 85,0 /1 mi de caracteres
87-11
Google logoGoogle
Gemini 3.8 Flash-Lite TTS
1055-15/152,307Sept 2026US$ 11,0 /1 mi de caracteres
98-12
ElevenLabs logoElevenLabs
v3 Conversational
1053-15/152,028Aug 2026US$ 50,0 /1 mi de caracteres
108-12
SpaceXAI logoSpaceXAI
SpaceXAI TTS
1052-14/142,652Mar 2026US$ 15,0 /1 mi de caracteres
118-12
Google logoGoogle
Gemini 3.8 Flash TTS
1051-15/152,294Sept 2026US$ 16,5 /1 mi de caracteres
1210-13
SpeechifyAI logoSpeechifyAI
Simba 3.0
1038-14/142,512Feb 2026US$ 6,6 /1 mi de caracteres
1312-18
BreezeBlue logoBreezeBlue
Breeze TTS 2Hugging FacePesos abertos
1025-15/152,021Aug 2026US$ 34,0 /1 mi de caracteres
1413-18
MiniMax logoMiniMax
Speech 2.8 HD
1023-14/143,181Feb 2026US$ 100,0 /1 mi de caracteres
1513-20
MiniMax logoMiniMax
Speech 2.8 Turbo
1022-14/143,308Feb 2026US$ 60,0 /1 mi de caracteres
1613-21
Fish Audio logoFish Audio
Fish Audio S2.1 Pro
1016-14/142,468Jun 2026US$ 15,0 /1 mi de caracteres
1713-21
Gradium logoGradium
Gradium TTS (Aug 2026)
1016-15/152,037Aug 2026US$ 47,2 /1 mi de caracteres
1813-22
Fish Audio logoFish Audio
OpenAudio S1 MiniHugging FacePesos abertos
1014-15/152,480Jun 2025US$ 15,0 /1 mi de caracteres
1915-22
Fish Audio logoFish Audio
Fish Audio S2 ProHugging FacePesos abertos
1009-13/133,724Mar 2026US$ 15,0 /1 mi de caracteres
2015-23
Mistral logoMistral
Voxtral TTSHugging FacePesos abertos
1008-14/142,432Mar 2026US$ 16,0 /1 mi de caracteres
2116-23
ElevenLabs logoElevenLabs
Multilingual v2
1005-14/143,182Aug 2023US$ 100,0 /1 mi de caracteres
2222
Fish Audio logoFish Audio
OpenAudio S1
10000/05,169Jun 2025US$ 15,0 /1 mi de caracteres
2319-24
ElevenLabs logoElevenLabs
Flash v2.5
995-14/143,193Dec 2024US$ 50,0 /1 mi de caracteres
2422-25
Fish Audio logoFish Audio
Fish Audio S1
990-14/143,266Oct 2025US$ 15,0 /1 mi de caracteres
2523-26
ElevenLabs logoElevenLabs
Turbo v2.5
980-15/152,544Jul 2024US$ 50,0 /1 mi de caracteres
2625-27
KugelAudio logoKugelAudio
Kugel 3
974-14/143,789Jun 2026N/D
2726-28
Boson AI logoBoson AI
Higgs Audio V3 TTSHugging FacePesos abertos
965-15/152,174Jun 2026US$ 15,0 /1 mi de caracteres
2827-31
Cartesia logoCartesia
Sonic 3
951-15/152,582Oct 2025US$ 49,0 /1 mi de caracteres
2928-33
NVIDIA logoNVIDIA
Magpie TTS ZeroshotPesos abertos
942-18/181,015Jul 2026N/D
3028-33
Krafton logoKrafton
Raon SpeechLM
941-15/152,005Apr 2026N/D
3128-33
Soniox logoSoniox
Soniox TTS Real-Time v2
940-16/161,955Aug 2026US$ 14,2 /1 mi de caracteres
3229-33
Resemble AI logoResemble AI
ChatterboxGitHubPesos abertos
936-15/152,561May 2025US$ 25,0 /1 mi de caracteres
3329-33
Noiz logoNoiz
Noiz TTS
935-15/152,080Nov 2025N/D
3434
Smallest.ai logoSmallest.ai
Lightning v3.1
898-16/162,183Mar 2026US$ 25,0 /1 mi de caracteres
3535
Coqui logoCoqui
846-17/172,347Nov 2023US$ 40,4 /1 mi de caracteres
3636
OpenVoice logoOpenVoice
OpenVoice v2Hugging FacePesos abertos
806-18/182,059Apr 2024US$ 8,3 /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.