AuK Technical Report: An Open-Source Foundational Model for Speech Generation and Editing

Benchmark Model Rank Results
speech-editing-on-ming-freeform-audio-edit-basicAuK#2ZH Deletion WER: 9.61EN Insertion WER: 3.47EN Deletion SIM: 0.83…
speech-editing-on-ming-freeform-audio-edit-basicAuK-Flash#3ZH Deletion WER: 9.85EN Insertion WER: 4.26EN Deletion SIM: 0.84…
speech-editing-on-speecheditbenchAuK#2Content: 91.83Emotion: 9.94Prosody: 71.33Paralinguistic: 38.5…
speech-editing-on-speecheditbenchAuK-Flash#3Content: 87.5Emotion: 6.29Prosody: 70.0Paralinguistic: 39.25…
text-to-speech-synthesis-on-seed-tts-test-enAuK#2WER: 1.02Speaker Similarity: 0.788
text-to-speech-synthesis-on-seed-tts-test-enAuK-Flash#3WER: 1.03Speaker Similarity: 0.772
text-to-speech-synthesis-on-seed-tts-test-hardAuK#2CER: 5.91Speaker Similarity: 0.782
text-to-speech-synthesis-on-seed-tts-test-hardAuK-Flash#4CER: 6.43Speaker Similarity: 0.784
text-to-speech-synthesis-on-seed-tts-test-zhAuK#5CER: 1.02Speaker Similarity: 0.814
text-to-speech-synthesis-on-seed-tts-test-zhAuK-Flash#6CER: 1.1Speaker Similarity: 0.814