Macで英語音声を文字起こしする手順｜OpenAI API完全マニュアル

mkdir -p ~/openai-transcribe
cd ~/openai-transcribe

export OPENAI_API_KEY="ここにあなたのAPIキー"

echo ${#OPENAI_API_KEY}

python3 -m venv .venv
source .venv/bin/activate

python3 -m pip install --upgrade pip
python3 -m pip install openai

open .

ls

nano split_wav.py

import wave

INPUT_FILE = "output.wav"
CHUNK_SECONDS = 700

with wave.open(INPUT_FILE, "rb") as wav:
    n_channels = wav.getnchannels()
    sampwidth = wav.getsampwidth()
    framerate = wav.getframerate()
    total_frames = wav.getnframes()

    frames_per_chunk = CHUNK_SECONDS * framerate
    total_chunks = (total_frames + frames_per_chunk - 1) // frames_per_chunk

    for i in range(total_chunks):
        out_name = f"part_{i+1}.wav"
        start_frame = i * frames_per_chunk
        wav.setpos(start_frame)
        frames = wav.readframes(frames_per_chunk)

        with wave.open(out_name, "wb") as out:
            out.setnchannels(n_channels)
            out.setsampwidth(sampwidth)
            out.setframerate(framerate)
            out.writeframes(frames)

        print(f"作成: {out_name}")

python3 split_wav.py
ls

nano transcribe_parts.py

from openai import OpenAI
import glob

client = OpenAI()

wav_files = sorted(glob.glob("part_*.wav"))
all_text = []

for wav_file in wav_files:
    print(f"文字起こし中: {wav_file}")
    with open(wav_file, "rb") as audio_file:
        result = client.audio.transcriptions.create(
            model="gpt-4o-transcribe",
            file=audio_file,
            language="en"
        )
    all_text.append(f"\n--- {wav_file} ---\n")
    all_text.append(result.text)

full_text = "\n".join(all_text)

with open("transcript.txt", "w", encoding="utf-8") as f:
    f.write(full_text)

print("完了: transcript.txt を保存しました")

python3 transcribe_parts.py

open transcript.txt

以下の英語文字起こしを、意味が通る自然な日本語に翻訳してください。
話し言葉の温度感は残しつつ、読みやすく整えてください。

cd ~/openai-transcribe
source .venv/bin/activate
export OPENAI_API_KEY="ここにあなたのAPIキー"

python3 split_wav.py
python3 transcribe_parts.py
open transcript.txt