step_audio_2
TiCo-StepAudio2-TiCoBench-v2-outputs
Step-Audio-2-mini outputs on TiCo-Bench-v2
Generated speech for all 2,000 TiCo-Bench-v2 items, from two models: the base
Step-Audio-2-mini and the TiCo version trained on it. Released so the duration
numbers in the paper can be recomputed from the audio rather than taken on
trust — every figure below is derived from these files and nothing else.
The two runs
directory
model
vanilla/
stepfun-ai/Step-Audio-2-mini, unmodified
tico-grpo-ckpt5400/
the same… See the full description on the dataset page: https://huggingface.co/datasets/ga642381/TiCo-StepAudio2-TiCoBench-v2-outputs.StepAudio2_Lu_Yin_BaiShiXi_TTSgenshin_StepAudio2_Lu_Yin_answer_Mavuika_audio_samples