本地部署SenseVoice的ONNX模型,识别音频并以字幕格式展示
1、创建虚拟环境并激活
2、安装sensevoice
pip install sensevoice-onnx -i https://pypi.tuna.tsinghua.edu.cn/simple
3、使用命令进行转换:
sensevoice --audio path/to/audio.wav --language auto --use_itn
language="auto", # "zh", "en", "yue", "ja", "ko", "nospeech"
如果模型下载失败,可以从hf镜像站下载:
https://hf-mirror.com/lovemefan/SenseVoice-onnx/tree/main
并手动拷贝至:
/home/zhangfanxp/miniconda3/envs/sense2/lib/python3.10/site-packages/sensevoice/resource/
https://hf-mirror.com/lovemefan/SenseVoice-onnx/tree/main
并手动拷贝至:
/home/zhangfanxp/miniconda3/envs/sense2/lib/python3.10/site-packages/sensevoice/resource/
另外:
whisper也支持按字幕格式输出。
评论
发表评论