某些音频识别得有问题

[hjj-lmx]

cf15582d-a84b-475b-bc80-385481f189b1.zip 掉了很多字

这是我调用得参数 res = self.sense_voice_model.generate( input=str(params["url"]), cache={}, language="auto", # "zn", "en", "yue", "ja", "ko", "nospeech" use_itn=True, batch_size_s=60, merge_vad=True, merge_length_s=15, ban_emo_unk=True ) text = rich_transcription_postprocess(res[0]["text"]) return text