Skip to content
ENTITY PROFILE / KNOWLEDGE GRAPH

Open TTS Leaderboard

RSS

Open TTS Leaderboard uses Qwen3 ASR for evaluating WER Open TTS Leaderboard uses H200 GPU to measure speed metrics Open TTS Leaderboard uses WavLM speaker embeddings for speaker similarity Open TTS Leaderboard ranks models by WER on Seed TTS Eval Open TTS Leaderboard ranks models by WER on CV3 Eval hexgrad/Kokoro-82M is top in English WER on Open TTS Leaderboard Supertone/supertonic-3 is top in English WER on Open TTS Leaderboard fisaudio/s2-pro is top in English WER on Open TTS Leaderboard k2-fsa/OmniVoice is a strong multilingual model evaluated by Open TTS Leaderboard FunAudioLLM/Fun-CosyVoice3-0.5B-2512 is a strong multilingual model evaluated by Open TTS Leaderboard bosonai/higgs-tts-3-4b is evaluated by Open TTS Leaderboard (shows improved WER with voice cloning) openbmb/VoxCPM2 is evaluated by Open TTS Leaderboard (shows improved WER with voice cloning) Open TTS Leaderboard will soon open-source its evaluation scripts, similar to Open ASR Leaderboard repo· Entity summary generated by Graphiti

RELATIONS
13Relations
CURRENT
13Current
EVENTS
0Events
REPORTS
1Source reports
KNOWLEDGE ATLAS0 entities · 0 relations
No public neighbors yet
Explore all published links of a node; previously discovered relations remain visible.

Relation timeline

BY VALID-FROM DATE
  1. Open TTS LeaderboardUSES GPU FOR SPEED EVALUATIONH200 GPU

    Open TTS Leaderboard uses H200 GPU to measure speed metrics

    Current1 sourceH200 GPU
  2. Supertone/supertonic-3TOP RANKED IN WEROpen TTS Leaderboard

    Supertone/supertonic-3 is top in English WER on Open TTS Leaderboard

    Current1 sourceSupertone/supertonic-3
  3. hexgrad/Kokoro-82MTOP RANKED IN WEROpen TTS Leaderboard

    hexgrad/Kokoro-82M is top in English WER on Open TTS Leaderboard

    Current1 sourcehexgrad/Kokoro-82M
  4. Open TTS LeaderboardUSES SPEAKER EMBEDDING MODELOpenClaw

    Open TTS Leaderboard uses WavLM speaker embeddings for speaker similarity

    Current1 sourceOpenClaw
  5. bosonai/higgs-tts-3-4bEVALUATED BYOpen TTS Leaderboard

    bosonai/higgs-tts-3-4b is evaluated by Open TTS Leaderboard (shows improved WER with voice cloning)

    Current1 sourcebosonai/higgs-tts-3-4b
  6. Open TTS LeaderboardUSES ASR MODEL FOR WERQwen3 ASR

    Open TTS Leaderboard uses Qwen3 ASR for evaluating WER

    Current1 sourceQwen3 ASR
  7. Open TTS LeaderboardRANKS BY WER ONCV3 Eval

    Open TTS Leaderboard ranks models by WER on CV3 Eval

    Current1 sourceCV3 Eval
  8. fisaudio/s2-proTOP RANKED IN WEROpen TTS Leaderboard

    fisaudio/s2-pro is top in English WER on Open TTS Leaderboard

    Current1 sourcefisaudio/s2-pro
  9. openbmb/VoxCPM2EVALUATED BYOpen TTS Leaderboard

    openbmb/VoxCPM2 is evaluated by Open TTS Leaderboard (shows improved WER with voice cloning)

    Current1 sourceopenbmb/VoxCPM2
  10. FunAudioLLM/Fun-CosyVoice3-0.5B-2512RANKED HIGH ON MULTILINGUAL PERFORMANCEOpen TTS Leaderboard

    FunAudioLLM/Fun-CosyVoice3-0.5B-2512 is a strong multilingual model evaluated by Open TTS Leaderboard

    Current1 sourceFunAudioLLM/Fun-CosyVoice3-0.5B-2512
  11. k2-fsa/OmniVoiceRANKED HIGH ON MULTILINGUAL PERFORMANCEOpen TTS Leaderboard

    k2-fsa/OmniVoice is a strong multilingual model evaluated by Open TTS Leaderboard

    Current1 sourcek2-fsa/OmniVoice
  12. Open TTS LeaderboardRANKS BY WER ONSeed TTS Eval

    Open TTS Leaderboard ranks models by WER on Seed TTS Eval

    Current1 sourceSeed TTS Eval
  13. Open TTS LeaderboardWILL OPEN SOURCE EVALUATION SCRIPTS LIKEOpenASR Leaderboard

    Open TTS Leaderboard will soon open-source its evaluation scripts, similar to Open ASR Leaderboard repo

    Current1 sourceOpenASR Leaderboard

Relations are facts Graphiti extracted from reports, not verbatim quotes. An end date means later reporting showed the relation no longer holds.

Source reports 1