Views
No views yet
1from transformers import pipeline
2from pathlib import Path
3
4pipe = pipeline(
5 "automatic-speech-recognition",
6 model="Parakeet-Inc/furigana_whisper_small_jsut",
7)
8
9def transcribe_with_prompt(pipe, audio_path: str | Path, prompt: str) -> str:
10 prompt_ids = pipe.tokenizer.get_prompt_ids(
11 prompt, return_tensors="pt"
12 ).to(pipe.device)
13 generate_kwargs = {"prompt_ids": prompt_ids}
14 result = pipe(str(audio_path), generate_kwargs=generate_kwargs)
15 return result["text"]
16
17# 実行例
18audio_path = "path/to/your/audio.wav"
19prompt = "明日は晴れ。"
20transcription = transcribe_with_prompt(pipe, audio_path, prompt)
21print(transcription) # アスワハレ。