21 lines
506 B
Python
21 lines
506 B
Python
|
from transformers import pipeline
|
||
|
|
|
||
|
|
transcriber = pipeline(
|
||
|
|
"automatic-speech-recognition",
|
||
|
|
model="./models--BELLE-2--Belle-whisper-large-v3-turbo-zh"
|
||
|
|
)
|
||
|
|
|
||
|
|
transcriber.model.config.forced_decoder_ids = (
|
||
|
|
transcriber.tokenizer.get_decoder_prompt_ids(
|
||
|
|
language="zh",
|
||
|
|
task="transcribe"
|
||
|
|
)
|
||
|
|
)
|
||
|
|
|
||
|
|
# 添加 return_timestamps=True
|
||
|
|
transcription = transcriber("audio.wav", return_timestamps=True)
|
||
|
|
print(transcription)
|
||
|
|
with open("audio.txt", "w", encoding="utf-8") as f:
|
||
|
|
f.write(str(transcription))
|
||
|
|
|