K-On! Houkago Live!! (PlayStation Portable) - The Cutting Room Floor

https://tcrf.net/K-On!_Houkago_Live!!_(PlayStation_Portable)

AI 语音翻译

import os
import json
import whisper
import time

# 加载Whisper模型
model = whisper.load_model("base")

# 待转录音频文件所在目录路径
directory = r"J:/2023-04-15_voices/haruhi_aac"

# 转录文本保存到JSON文件中的路径和文件名
json_file = 'haruhi_transcriptions.json'

# 如果JSON文件存在，则读取其中的内容
if os.path.exists(json_file):
    with open(json_file, 'r', encoding='utf-8') as f:
        transcription_dict = json.load(f)
else:
    # 如果JSON文件不存在，则创建一个空字典
    transcription_dict = {}

# 统计待转录音频文件总数
total_files = 0
for filename in os.listdir(directory):
    if filename.endswith(".m4a") and filename not in transcription_dict:
        total_files += 1

# 定义一个函数，用于保存转录结果到JSON文件中
def save_transcriptions(transcription_dict, json_file):
    with open(json_file, 'w', encoding='utf-8') as f:
        json.dump(transcription_dict, f, ensure_ascii=False, indent=2)
    print(f"已保存转录结果到{json_file}")

if total_files == 0:
    print("没有需要转录的文件")
else:
    # 记录已经转录的文件数量和单次处理时间的列表
    count = 0
    process_times = []

    # 遍历待转录音频文件
    for filename in os.listdir(directory):

        # 只处理.m4a格式的音频文件
        if filename.endswith(".m4a") and filename not in transcription_dict:

            # 进行转录并将结果添加到字典中
            filepath = os.path.join(directory, filename)
            start_time = time.time()
            result = model.transcribe(filepath, language='Japanese')
            process_time = time.time() - start_time
            process_times.append(process_time)
            transcription_text = result["text"]
            transcription_dict[filename] = transcription_text

            # 增加转录文件计数器的值
            count += 1

            # 计算剩余文件数量和平均单次处理时间
            remaining_files = total_files - count
            average_time = sum(process_times) / len(process_times)
            remaining_time = remaining_files * average_time

            # 打印出剩余时间、剩余文件数量、文件名和转录文本
            print(f"---------------------------------")
            print(f"文件名：{filename}", transcription_text)
            print(f"本次：{process_time:.2f}秒、平均：{average_time:.2f}秒")
            print(f"还剩 {remaining_files} 个文件、预计总时长：{remaining_time:.2f}秒")

            # 每10个文件保存一次转录结果到JSON文件中
            if count % 10 == 0:
                save_transcriptions(transcription_dict, json_file)

    # 最后再保存一次转录结果到JSON文件中
    save_transcriptions(transcription_dict, json_file)