OmniVoice: Towards Omnilingual Zero-Shot Text-to-Speech with Diffusion Language Models

Benchmark Model Rank Results
text-to-speech-synthesis-on-librispeech-pc-test-cleanCosyVoice3#2WER: 1.59UTMOS: 4.28Speaker Similarity: 0.694
text-to-speech-synthesis-on-librispeech-pc-test-cleanQwen3-TTS#3WER: 1.6UTMOS: 4.41Speaker Similarity: 0.704
text-to-speech-synthesis-on-librispeech-pc-test-cleanVoxCPM#5WER: 1.74UTMOS: 4.18Speaker Similarity: 0.717
text-to-speech-synthesis-on-librispeech-pc-test-cleanIndexTTS2#8WER: 2.35UTMOS: 4.06Speaker Similarity: 0.700
text-to-speech-synthesis-on-seed-tts-test-zhOmniVoice#3CER: 0.84Speaker Similarity: 0.777
text-to-speech-synthesis-on-seed-tts-test-zhCosyVoice3#9CER: 1.14Speaker Similarity: 0.778