TTS Arena — AI Voice Model Leaderboard

@ action: inmenu

QPrintPreviewDialog QDialogButtonBox QShortcut QDialogButtonBox QPrintPreviewDialog

KCharselect unicode block name

A fair, community-driven way to evaluate AI voice models

KCharselect unicode block name

Standardized evaluation metrics ciki har da MOS (Mean Opinion Score), karamin adadin kuskure, masu magana da yawa, da kuma real-time factor a kan dukkan 20+ models.

QShortcut

QDialogButtonBox

QPrintPreviewDialog

Yi halitta ga rubutun da ya dace da nau'ikan biyu daban-daban kuma ka kwatanta ingancin sauti, dabi'a, da sauri kai tsaye cikin mai bincike.

20+ Models Ranked

Duk wani nau'i a kan TTS.ai an yi shi da kuma rarraba. Cire ta hanyar gudu, inganci, goyon bayan harshe, halaye, da lasisin don samun nau'inka na musamman.

@ action

Ka yi zurfi cikin aikin kowace siffar: jinkiri, gudu, amfani da VRAM, harsuna masu goyon baya, ingancin kwaikwayo, da kuma sakamakon ma'aunin jin dadi.

QDialogButtonBox

Browse the leaderboard, kwatanta models, da kuri'a a kan ingancin - dukansu kyauta. Babu asusun da ake buƙata don bincika matsayi da kuma benchmarks.

QPrintPreviewDialog

All 20+ models compete head-to-head for the top ranking

KokoroKokoro

Free

Lightweight 82M parameter model delivering studio-quality speech with blazing-fast inference.

Fast 5/5

Mafi kyawun ga: Top-ranked free model — best speed-to-quality ratio on the leaderboard

QDialogButtonBox Kokoro

ChatterboxChatterbox

Premium

State-of-the-art zero-shot voice cloning with emotion control from Resemble AI.

Medium 5/5 QShortcut

Mafi kyawun ga: Highest-rated voice cloning model with emotion control capabilities

QDialogButtonBox Chatterbox

CosyVoice 2CosyVoice 2

Standard

Alibaba's scalable streaming TTS with human-parity naturalness and near-zero latency.

Medium 5/5 QShortcut

Mafi kyawun ga: Top multilingual model with human-parity naturalness scores

QDialogButtonBox CosyVoice 2

StyleTTS 2StyleTTS 2

Premium

Human-level text-to-speech through style diffusion and adversarial training.

Medium 5/5

Mafi kyawun ga: Highest single-speaker MOS score among all open-source models

QDialogButtonBox StyleTTS 2

Sesame CSMSesame CSM

Premium

Conversational speech model generating natural dialogue with appropriate timing and emotion.

Slow 5/5

Mafi kyawun ga: QShortcut

QDialogButtonBox Sesame CSM

Yadda TTS Arena ke aiki

Zaɓi ingancin magana kuma ka taimaka wajen tantance mafi kyawun sifofin AI

1

QShortcut

Kallon duk 20 + siffofin da aka rarraba ta inganci, sauri, da halaye. Cire ta matakin (farashi, siffar, premium) ko ayyuka na musamman.

2

@ action

Ka zaɓa nau'i biyu kuma ka samar da rubutu guda tare da su biyu. Ka saurari fitarwa kuma ka kwatanta dabi'a, haske, da bayyanar ra'ayi.

3

QShortcut

Bayan da ka kwatanta, za ka iya kada kuri'a game da nau'in da ya fi kyau. Kuri'un ka na taimakawa wajen tantance jama'a da kuma taimaka wa sauran masu amfani da su zaɓa.

4

@ action

Yi amfani da bayanai na leaderboard da kuma ra'ayoyin jama'a don zaɓar mafi kyawun sigar don amfani da ka'idojinka, kasafin kuɗi, da kuma bukatun inganci.

Mẽne ne TTS Arena?

A community-driven approach to ranking AI voice models

QPrintPreviewDialog

@ action: inmenu

  • Text same, two anonymous models
  • An bayyana sunayen ma'aurata bayan zaben
  • Nau'i-nau'i masu ban sha'awa a kowace motsi
  • QShortcut

KCharselect unicode block name

An ƙayyade nau'ukan ta hanyar amfani da tsarin ƙididdiga na Elo, irin algorithm da ake amfani da shi wajen ƙayyade 'yan wasa na chess. Yin nasara kan nau'i mai girman ƙididdiga yana ba da dama fiye da yin nasara kan nau'i mai ƙarancin ƙididdiga. A kan dubban kuri'u, wannan na samar da ƙididdiga mai aminci wanda ke nuna fifikon jama'a na gaskiya.

  • Algorithm na rarraba Elo-da-dacewa
  • QShortcut
  • Statistical trust intervals
  • QShortcut

Preview available

Yadda mu 20 + models kwatanta a kan muhimmanci bambance-bambance

@ action DakataEthiopian month 11 - LongNamePossessive QPrintPreviewDialog QSoftKeyManager @ item Spelling dictionary @ action
Kokoro QDialogButtonBox 4.5/5 QPrintPreviewDialog 8
Bark @ action 4.0/5 Media 13
CosyVoice2 @ action 4.5/5 Media 6
Tortoise TTS PremiumLanguage 4.8/5 QPrintPreviewDialog 1
Chatterbox PremiumLanguage 4.7/5 Media 1
StyleTTS 2 PremiumLanguage 4.7/5 QPrintPreviewDialog 1

QSql

Me ya sa wani nau'in TTS ya fi girma a cikin filin

QFontDatabase

Shin yana ji kamar mutum ne na gaskiya? Prosody na dabi'a, rikodi, da nau'ikan intonation waɗanda suke daidaita maganar mutum. Babu robotic artifacts ko tsawo ba na dabi'a ba.

QFontDatabase

Shin sauti yana isar da ra'ayi mai dacewa da kuma mai da hankali? Mafi kyawun ma'aurata suna kula da tambayoyi, kiraye-kiraye, da kuma yanayin ra'ayi na dabi'a.

QPrintPreviewDialog

Shin yana fassara kowace kalma daidai? Yana kula da kalmomi masu ban sha'awa, lambobi, siffofi, da sunaye na waje ba tare da kuskure ba ko kuma sauti masu kama da na almara.

QDialogButtonBox

Zaɓuɓɓukanka suna da tasiri kai tsaye kan teburin jagoranci. Duk wani kwatanta yana taimaka wa al'umma wajen gano mafi kyawun ma'aurata.

Ka shiga cikin TTS Arena

Tambayar da ake yi da yawa

Tambayoyi masu yawan gaske game da TTS Arena da dabaru

The TTS Arena ne a leaderboard da kwatanta kayan aiki ga AI rubutu-to- magana models. It ranks 20+ models based on official benchmarks and community votes, helping users find the best model for their needs through standardized evaluation and side-by-side comparison.

An yi la'akari da nau'ikan a kan matakan da dama: MOS (Mean Opinion Score) domin ingancin mutum, adadin kuskure na alamomin don daidaitaccen magana, ma'aunin lokaci na gaskiya don sauri, amfani da VRAM don inganci, da kuri'u na al'umma don fifikon duniyar gaskiya. An yi la'akari da kuri'u don samar da matsayi na gabaɗaya.

@ label

Rankings depend on criteria. Kokoro leads in speed-to-quality ratio. StyleTTS 2 achieves the highest single-speaker MOS. Chatterbox tops voice cloning rankings. CosyVoice 2 leads multilingual quality. Check the leaderboard for current standings in each category.

Yes. Listen to side-by-side comparisons and vote for the model that sounds better. Voting is free and does not require a account. Community votes directly influence the rankings and help surface the best models for different use cases.

An sabunta lissafin lissafi na aikace-aikace lokacin da aka ƙara sabon nau'i ko nau'ikan da ke akwai suna karɓar sabuntawa masu mahimmanci. Rankings na jama'a suna sabuntawa a cikin lokaci na gaskiya kamar yadda kuri'u ke zuwa. Muna sake nazarin dukkan nau'ikan kowane wata don tabbatar da daidaituwa da daidaitacce.

@ action

Ka shigar da misalin rubutu, ka zaɓi nau'ikan biyu, kuma ka danna samar. Nau'ikan biyu suna samar da sauti daga rubutun guda. Ka saurari nasararsu biyu kuma ka yanke hukunci wanda yake da sauti mafi dabi'a, bayyananne, da bayyanawa. Za ka iya zaɓar nau'in da kake so daga nan.

Na'am. Mun wallafa hanyarmu ta binciki, kalmomin gwaji, da kuma sharuɗɗan nazari. An gwada dukkan sifofi a ƙarƙashin halaye masu kama da juna a kan na'urar GPU mai kama da juna. Ƙungiyoyin jama'a na iya samar da sakamakon ta amfani da jerin gwaje-gwajenmu da aka wallafa da kuma sharuɗɗan sakamako.

The arena ya mayar da hankali a kan 20+ bude-source models da aka shirya a kan TTS.ai. Ba mu daidaita sabis na kasuwanci kamar ElevenLabs ko Google TTS, amma MOS scores da metrics suna da alaƙa da buga benchmarks daga waɗannan sabis.

@ action

Kokoro (farashin) ya samu sakamako na inganci 5/5, wanda ya haɗu da nau'ikan premium da yawa. Amfanin farko na nau'ikan premium sune halaye masu dacewa kamar clone na magana (Chatterbox), style diffusion (StyleTTS 2), da maganar hira (Sesame CSM) fiye da ingancin sauti na asali.
5.0/5 (1)

@ info

Ka yi Zabe a cikin TTS Arena

Ku saurari muryoyin AI, ku kada kuri'a don mafi kyawun, kuma ku bincika jerinmu na 20+ da jama'a suka jagoranta.