ggerganov / whisper.cpp

Port of OpenAI's Whisper model in C/C++
MIT License
35.38k stars 3.61k forks source link

Stream: noise ouput #2104

Open YutaoChow opened 6 months ago

YutaoChow commented 6 months ago

When I attempted to transcribe the audio into Chinese text, it generated a lot of noise, despite having set both the microphone and speaker volumes to zero.

My Computer Config:

image image

Shell Command: ./stream -m ./models/ggml-base.bin --step 0 --length 30000 -vth 0.6 --language zh --capture 1

Shell Output: image

Can anyone help me,thanks!

Huangdahei commented 1 month ago

麦克风没有音频输入的时候会有这种输出,所以可以考虑设置声音振幅阈值来过滤掉无音频输入时的输出。