pythonのwisperを使用して自動で文字起こし＆テロップ作成のプログラムを公開

conda install -c conda-forge ffmpeg

ffmpeg -version

pip install -U openai-whisper

import whisper

mov_path = "動画のパスを入力"
model = whisper.load_model("base")
result = model.transcribe(mov_path)

1
00:00:00,000 --> 00:00:00,814
どうもとしです。

2
00:00:00,814 --> 00:00:04,071
今回AI×Python

3
00:00:04,071 --> 00:00:04,885
自動で動画からテロップを作成という

7
00:00:04,885 --> 00:00:05,700
テーマで話をしていきたいと思います

8
00:00:05,700 --> 00:00:07,346
AIっていうものを使うことによって

def split_text_by_punctuation_and_length(text, max_length=27):
    # 句読点、感嘆符、スペースで分割
    sentences = re.split(r'(?<=[。、！？!?\s])', text)
    result = []
    buffer = ""
    for sentence in sentences:
        for char in sentence:
            buffer += char
            # バッファの文字数が最大値を超えた場合、次に移行
            if len(buffer) >= max_length:
                result.append(buffer)
                buffer = ""
        # 文章単位の句読点で分割
        if buffer:
            result.append(buffer)
            buffer = ""
    return result

def format_time(seconds):
    """秒をSRTフォーマットのタイムスタンプに変換"""
    millis = int((seconds % 1) * 1000)
    seconds = int(seconds)
    hours = seconds // 3600
    minutes = (seconds % 3600) // 60
    seconds = seconds % 60
    return f"{hours:02}:{minutes:02}:{seconds:02},{millis:03}"

def write_srt(segments, srt_path):
    """SRTファイルに書き出し"""
    with open(srt_path, "w", encoding="utf-8") as srt_file:
        for i, (start, end, text) in enumerate(segments):
            srt_file.write(f"{i + 1}\n")
            srt_file.write(f"{format_time(start)} --> {format_time(end)}\n")
            srt_file.write(f"{text}\n\n")

def transcribe_with_custom_split(result, out_srt_path):
    # 出力セグメントを処理して分割
    max_chars = 27
    segments = result['segments']
    formatted_segments = []

    for segment in segments:
        start = segment["start"]
        end = segment["end"]
        text = segment["text"]

        # 一定文字数ごとに句読点,スペースで分割
        split_texts = split_text_by_punctuation_and_length(text, max_chars)

        # 各分割部分に新しいタイムスタンプを割り当て
        duration = (end - start) / len(split_texts)
        for i, sub_text in enumerate(split_texts):
            sub_start = start + i * duration
            sub_end = sub_start + duration
            formatted_segments.append((sub_start, sub_end, sub_text))
    # SRTファイルに書き出し
    write_srt(formatted_segments, out_srt_path)

import os
import re
import whisper


def get_file_path():
    movie_file_path = input('今回読み込ませるフルパスを入力してください\nmovie file name:').replace('"', '')
    file_name = os.path.splitext(os.path.basename(movie_file_path))[0]
    out_srt_path = f'{file_name}.srt'
    return movie_file_path, out_srt_path


def format_time(seconds):
    """秒をSRTフォーマットのタイムスタンプに変換"""
    millis = int((seconds % 1) * 1000)
    seconds = int(seconds)
    hours = seconds // 3600
    minutes = (seconds % 3600) // 60
    seconds = seconds % 60
    return f"{hours:02}:{minutes:02}:{seconds:02},{millis:03}"


def split_text_by_punctuation_and_length(text, max_length=27):
    # 句読点,でまず分割
    sentences = re.split(r'(?<=[。、！？!?\s])', text)
    result = []
    buffer = ""
    for sentence in sentences:
        for char in sentence:
            buffer += char
            # バッファの文字数が最大値を超えた場合、次に移行
            if len(buffer) >= max_length:
                result.append(buffer)
                buffer = ""
        # 文章単位の句読点で分割
        if buffer:
            result.append(buffer)
            buffer = ""
    return result


def write_srt(segments, srt_path):
    """SRTファイルに書き出し"""
    with open(srt_path, "w", encoding="utf-8") as srt_file:
        for i, (start, end, text) in enumerate(segments):
            srt_file.write(f"{i + 1}\n")
            srt_file.write(f"{format_time(start)} --> {format_time(end)}\n")
            srt_file.write(f"{text}\n\n")


def transcribe_with_custom_split(result, out_srt_path):
    # 出力セグメントを処理して分割
    max_chars = 27
    segments = result['segments']
    formatted_segments = []

    for segment in segments:
        start = segment["start"]
        end = segment["end"]
        text = segment["text"]

        # 一定文字数ごとに句読点,スペースで分割
        split_texts = split_text_by_punctuation_and_length(text, max_chars)

        # 各分割部分に新しいタイムスタンプを割り当て
        duration = (end - start) / len(split_texts)
        for i, sub_text in enumerate(split_texts):
            sub_start = start + i * duration
            sub_end = sub_start + duration
            formatted_segments.append((sub_start, sub_end, sub_text))
    # SRTファイルに書き出し
    write_srt(formatted_segments, out_srt_path)


mov_path, srt_path = get_file_path()
model = whisper.load_model("base")
split_result = model.transcribe(mov_path)
transcribe_with_custom_split(split_result, srt_path)
print('SRTファイルの書き出しが完了しました')