feat(task): 重构视频剪辑流程并添加新功能

-重构了 start_subclip 函数，优化了视频剪辑流程
- 新增 clip_video 函数，用于裁剪视频片段
- 改进了字幕生成和处理逻辑- 优化了音频合并和处理
- 更新了任务状态管理

app/services/audio_merger.py

@@ -46,7 +46,7 @@ def merge_audio_files(task_id: str, audio_files: list, total_duration: float, li
             tts_audio = AudioSegment.from_file(audio_file)
 
             # 获取片段的开始和结束时间
-            start_time, end_time = segment['new_timestamp'].split('-')
+            start_time, end_time = segment['timestamp'].split('-')
             start_seconds = utils.time_to_seconds(start_time)
             end_seconds = utils.time_to_seconds(end_time)
 

app/services/clip_video.py

@@ -0,0 +1,227 @@
+#!/usr/bin/env python
+# -*- coding: UTF-8 -*-
+
+'''
+@Project: NarratoAI
+@File   : clip_video
+@Author : 小林同学
+@Date   : 2025/5/6 下午6:14 
+'''
+
+import os
+import subprocess
+import json
+import hashlib
+import logging
+from typing import Dict, List, Optional
+from pathlib import Path
+
+# 配置日志
+logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(name)s - %(levelname)s - %(message)s')
+logger = logging.getLogger(__name__)
+
+
+def parse_timestamp(timestamp: str) -> tuple:
+    """
+    解析时间戳字符串，返回开始和结束时间
+    
+    Args:
+        timestamp: 格式为'HH:MM:SS-HH:MM:SS'的时间戳字符串
+        
+    Returns:
+        tuple: (开始时间, 结束时间) 格式为'HH:MM:SS'
+    """
+    start_time, end_time = timestamp.split('-')
+    return start_time, end_time
+
+
+def calculate_end_time(start_time: str, duration: float, extra_seconds: float = 1.0) -> str:
+    """
+    根据开始时间和持续时间计算结束时间
+    
+    Args:
+        start_time: 开始时间，格式为'HH:MM:SS'
+        duration: 持续时间，单位为秒
+        extra_seconds: 额外添加的秒数，默认为1秒
+        
+    Returns:
+        str: 计算后的结束时间，格式为'HH:MM:SS'
+    """
+    h, m, s = map(int, start_time.split(':'))
+    total_seconds = h * 3600 + m * 60 + s + duration + extra_seconds
+
+    h_new = int(total_seconds // 3600)
+    m_new = int((total_seconds % 3600) // 60)
+    s_new = int(total_seconds % 60)
+
+    return f"{h_new:02d}:{m_new:02d}:{s_new:02d}"
+
+
+def check_hardware_acceleration() -> Optional[str]:
+    """
+    检查系统支持的硬件加速选项
+    
+    Returns:
+        Optional[str]: 硬件加速参数，如果不支持则返回None
+    """
+    # 检查NVIDIA GPU支持
+    try:
+        nvidia_check = subprocess.run(
+            ["ffmpeg", "-hwaccel", "cuda", "-i", "/dev/null", "-f", "null", "-"],
+            stderr=subprocess.PIPE, stdout=subprocess.PIPE, text=True, check=False
+        )
+        if nvidia_check.returncode == 0:
+            return "cuda"
+    except Exception:
+        pass
+
+    # 检查MacOS videotoolbox支持
+    try:
+        videotoolbox_check = subprocess.run(
+            ["ffmpeg", "-hwaccel", "videotoolbox", "-i", "/dev/null", "-f", "null", "-"],
+            stderr=subprocess.PIPE, stdout=subprocess.PIPE, text=True, check=False
+        )
+        if videotoolbox_check.returncode == 0:
+            return "videotoolbox"
+    except Exception:
+        pass
+
+    # 检查Intel Quick Sync支持
+    try:
+        qsv_check = subprocess.run(
+            ["ffmpeg", "-hwaccel", "qsv", "-i", "/dev/null", "-f", "null", "-"],
+            stderr=subprocess.PIPE, stdout=subprocess.PIPE, text=True, check=False
+        )
+        if qsv_check.returncode == 0:
+            return "qsv"
+    except Exception:
+        pass
+
+    return None
+
+
+def clip_video(
+        video_origin_path: str,
+        tts_result: List[Dict],
+        output_dir: Optional[str] = None,
+        task_id: Optional[str] = None
+) -> Dict[str, str]:
+    """
+    根据时间戳裁剪视频
+    
+    Args:
+        video_origin_path: 原始视频的路径
+        tts_result: 包含时间戳和持续时间信息的列表
+        output_dir: 输出目录路径，默认为None时会自动生成
+        task_id: 任务ID，用于生成唯一的输出目录，默认为None时会自动生成
+        
+    Returns:
+        Dict[str, str]: 时间戳到裁剪后视频路径的映射
+    """
+    # 检查视频文件是否存在
+    if not os.path.exists(video_origin_path):
+        raise FileNotFoundError(f"视频文件不存在: {video_origin_path}")
+
+    # 如果未提供task_id，则根据输入生成一个唯一ID
+    if task_id is None:
+        content_for_hash = f"{video_origin_path}_{json.dumps(tts_result)}"
+        task_id = hashlib.md5(content_for_hash.encode()).hexdigest()
+
+    # 设置输出目录
+    if output_dir is None:
+        output_dir = os.path.join(
+            os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__)))),
+            "storage", "temp", "clip_video", task_id
+        )
+
+    # 确保输出目录存在
+    Path(output_dir).mkdir(parents=True, exist_ok=True)
+
+    # 检查硬件加速支持
+    hwaccel = check_hardware_acceleration()
+    hwaccel_args = []
+    if hwaccel:
+        hwaccel_args = ["-hwaccel", hwaccel]
+        logger.info(f"使用硬件加速: {hwaccel}")
+
+    # 存储裁剪结果
+    result = {}
+
+    for item in tts_result:
+        timestamp = item["timestamp"]
+        start_time, _ = parse_timestamp(timestamp)
+
+        # 根据持续时间计算真正的结束时间（加上1秒余量）
+        duration = item["duration"]
+        calculated_end_time = calculate_end_time(start_time, duration)
+
+        # 格式化输出文件名
+        output_filename = f"vid-{start_time.replace(':', '-')}-{calculated_end_time.replace(':', '-')}.mp4"
+        output_path = os.path.join(output_dir, output_filename)
+
+        # 构建FFmpeg命令
+        ffmpeg_cmd = [
+            "ffmpeg", "-y", *hwaccel_args,
+            "-i", video_origin_path,
+            "-ss", start_time,
+            "-to", calculated_end_time,
+            "-c:v", "h264_videotoolbox" if hwaccel == "videotoolbox" else "libx264",
+            "-c:a", "aac",
+            "-strict", "experimental",
+            output_path
+        ]
+
+        # 执行FFmpeg命令
+        try:
+            logger.info(f"裁剪视频片段: {timestamp} -> {start_time}到{calculated_end_time}")
+            logger.debug(f"执行命令: {' '.join(ffmpeg_cmd)}")
+
+            process = subprocess.run(
+                ffmpeg_cmd,
+                stdout=subprocess.PIPE,
+                stderr=subprocess.PIPE,
+                text=True,
+                check=True
+            )
+
+            result[timestamp] = output_path
+            logger.info(f"成功裁剪视频片段: {timestamp} -> {output_path}")
+
+        except subprocess.CalledProcessError as e:
+            logger.error(f"裁剪视频片段失败: {timestamp}")
+            logger.error(f"错误信息: {e.stderr}")
+            raise RuntimeError(f"视频裁剪失败: {e.stderr}")
+
+    return result
+
+
+if __name__ == "__main__":
+    video_origin_path = "/Users/apple/Desktop/home/NarratoAI/resource/videos/qyn2-2无片头片尾.mp4"
+
+    tts_result = [{'timestamp': '00:00:00-00:01:15',
+                   'audio_file': '/Users/apple/Desktop/home/NarratoAI/storage/tasks/qyn2-2-demo/audio_00_00_00-00_01_15.mp3',
+                   'subtitle_file': '/Users/apple/Desktop/home/NarratoAI/storage/tasks/qyn2-2-demo/subtitle_00_00_00-00_01_15.srt',
+                   'duration': 25.55,
+                   'text': '好的各位，欢迎回到我的频道！《庆余年 2》刚开播就给了我们一个王炸！范闲在北齐"死"了？这怎么可能！上集片尾那个巨大的悬念，这一集就立刻揭晓了！范闲假死归来，他面临的第一个，也是最大的难关，就是如何面对他最敬爱的，同时也是最可怕的那个人——庆帝！'},
+                  {'timestamp': '00:01:15-00:04:40',
+                   'audio_file': '/Users/apple/Desktop/home/NarratoAI/storage/tasks/qyn2-2-demo/audio_00_01_15-00_04_40.mp3',
+                   'subtitle_file': '/Users/apple/Desktop/home/NarratoAI/storage/tasks/qyn2-2-demo/subtitle_00_01_15-00_04_40.srt',
+                   'duration': 13.488,
+                   'text': '但我们都知道，他绝不可能就这么轻易退场！第二集一开场，范闲就已经秘密回到了京都。他的生死传闻，可不像我们想象中那样只是小范围流传，而是…'},
+                  {'timestamp': '00:04:58-00:05:45',
+                   'audio_file': '/Users/apple/Desktop/home/NarratoAI/storage/tasks/qyn2-2-demo/audio_00_04_58-00_05_45.mp3',
+                   'subtitle_file': '/Users/apple/Desktop/home/NarratoAI/storage/tasks/qyn2-2-demo/subtitle_00_04_58-00_05_45.srt',
+                   'duration': 21.363,
+                   'text': '"欺君之罪"！在封建王朝，这可是抄家灭族的大罪！搁一般人，肯定脚底抹油溜之大吉了。但范闲是谁啊？他偏要反其道而行之！他竟然决定，直接去见庆帝！冒着天大的风险，用"假死"这个事实去赌庆帝的态度！'},
+                  {'timestamp': '00:05:45-00:06:00',
+                   'audio_file': '/Users/apple/Desktop/home/NarratoAI/storage/tasks/qyn2-2-demo/audio_00_05_45-00_06_00.mp3',
+                   'subtitle_file': '/Users/apple/Desktop/home/NarratoAI/storage/tasks/qyn2-2-demo/subtitle_00_05_45-00_06_00.srt',
+                   'duration': 7.675, 'text': '但想见庆帝，哪有那么容易？范闲艺高人胆大，竟然选择了最激进的方式——闯宫！'}]
+
+    # 使用方法示例
+    try:
+        result = clip_video(video_origin_path, tts_result)
+        print("裁剪结果:")
+        print(json.dumps(result, indent=4, ensure_ascii=False))
+    except Exception as e:
+        print(f"发生错误: {e}")

app/services/merger_video.py

@@ -0,0 +1,543 @@
+#!/usr/bin/env python
+# -*- coding: UTF-8 -*-
+
+'''
+@Project: NarratoAI
+@File   : merger_video
+@Author : 小林同学
+@Date   : 2025/5/6 下午7:38 
+'''
+
+import os
+import subprocess
+import logging
+from enum import Enum
+from typing import List, Optional, Tuple, Dict, Any
+import shutil
+
+# 设置日志
+logger = logging.getLogger(__name__)
+
+
+class VideoAspect(Enum):
+    """视频宽高比枚举"""
+    portrait = "portrait"  # 竖屏 9:16
+    landscape = "landscape"  # 横屏 16:9
+    square = "square"  # 方形 1:1
+
+    def to_resolution(self) -> Tuple[int, int]:
+        """根据宽高比返回标准分辨率"""
+        if self == VideoAspect.portrait:
+            return 1080, 1920  # 竖屏 9:16
+        elif self == VideoAspect.landscape:
+            return 1920, 1080  # 横屏 16:9
+        elif self == VideoAspect.square:
+            return 1080, 1080  # 方形 1:1
+        else:
+            return 1080, 1920  # 默认竖屏
+
+
+def check_ffmpeg_installation() -> bool:
+    """
+    检查ffmpeg是否已安装
+    
+    Returns:
+        bool: 如果安装则返回True，否则返回False
+    """
+    try:
+        subprocess.run(['ffmpeg', '-version'], stdout=subprocess.PIPE, stderr=subprocess.PIPE, check=True)
+        return True
+    except (subprocess.SubprocessError, FileNotFoundError):
+        logger.error("ffmpeg未安装或不在系统PATH中，请安装ffmpeg")
+        return False
+
+
+def get_hardware_acceleration_option() -> Optional[str]:
+    """
+    根据系统环境选择合适的硬件加速选项
+    
+    Returns:
+        Optional[str]: 硬件加速参数，如果不支持则返回None
+    """
+    try:
+        # 检查NVIDIA GPU支持
+        nvidia_check = subprocess.run(
+            ['ffmpeg', '-hide_banner', '-hwaccels'],
+            stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True
+        )
+        output = nvidia_check.stdout.lower()
+
+        if 'cuda' in output:
+            return 'cuda'
+        elif 'nvenc' in output:
+            return 'nvenc'
+        elif 'qsv' in output:  # Intel Quick Sync
+            return 'qsv'
+        elif 'videotoolbox' in output:  # macOS
+            return 'videotoolbox'
+        elif 'vaapi' in output:  # Linux VA-API
+            return 'vaapi'
+        else:
+            logger.info("没有找到支持的硬件加速器，将使用软件编码")
+            return None
+    except Exception as e:
+        logger.warning(f"检测硬件加速器时出错: {str(e)}，将使用软件编码")
+        return None
+
+
+def check_video_has_audio(video_path: str) -> bool:
+    """
+    检查视频是否包含音频流
+    
+    Args:
+        video_path: 视频文件路径
+        
+    Returns:
+        bool: 如果视频包含音频流则返回True，否则返回False
+    """
+    if not os.path.exists(video_path):
+        logger.warning(f"视频文件不存在: {video_path}")
+        return False
+        
+    probe_cmd = [
+        'ffprobe', '-v', 'error', 
+        '-select_streams', 'a:0', 
+        '-show_entries', 'stream=codec_type', 
+        '-of', 'csv=p=0', 
+        video_path
+    ]
+    
+    try:
+        result = subprocess.run(probe_cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True, check=False)
+        return result.stdout.strip() == 'audio'
+    except Exception as e:
+        logger.warning(f"检测视频音频流时出错: {str(e)}")
+        return False
+
+
+def create_ffmpeg_concat_file(video_paths: List[str], concat_file_path: str) -> str:
+    """
+    创建ffmpeg合并所需的concat文件
+    
+    Args:
+        video_paths: 需要合并的视频文件路径列表
+        concat_file_path: concat文件的输出路径
+        
+    Returns:
+        str: concat文件的路径
+    """
+    with open(concat_file_path, 'w', encoding='utf-8') as f:
+        for video_path in video_paths:
+            # 使用绝对路径并转义特殊字符
+            abs_path = os.path.abspath(video_path).replace('\\', '\\\\').replace(':', '\\:')
+            f.write(f"file '{abs_path}'\n")
+    return concat_file_path
+
+
+def process_single_video(
+        input_path: str,
+        output_path: str,
+        target_width: int,
+        target_height: int,
+        keep_audio: bool = True,
+        hwaccel: Optional[str] = None
+) -> str:
+    """
+    处理单个视频：调整分辨率、帧率等
+    
+    Args:
+        input_path: 输入视频路径
+        output_path: 输出视频路径
+        target_width: 目标宽度
+        target_height: 目标高度
+        keep_audio: 是否保留音频
+        hwaccel: 硬件加速选项
+        
+    Returns:
+        str: 处理后的视频路径
+    """
+    if not os.path.exists(input_path):
+        raise FileNotFoundError(f"找不到视频文件: {input_path}")
+
+    # 构建基本命令
+    command = ['ffmpeg', '-y']
+
+    # 添加硬件加速参数
+    if hwaccel:
+        if hwaccel == 'cuda' or hwaccel == 'nvenc':
+            command.extend(['-hwaccel', 'cuda'])
+        elif hwaccel == 'qsv':
+            command.extend(['-hwaccel', 'qsv'])
+        elif hwaccel == 'videotoolbox':
+            command.extend(['-hwaccel', 'videotoolbox'])
+        elif hwaccel == 'vaapi':
+            command.extend(['-hwaccel', 'vaapi', '-vaapi_device', '/dev/dri/renderD128'])
+
+    # 输入文件
+    command.extend(['-i', input_path])
+
+    # 处理音频
+    if not keep_audio:
+        command.extend(['-an'])  # 移除音频
+    else:
+        # 检查输入视频是否有音频流
+        has_audio = check_video_has_audio(input_path)
+        if has_audio:
+            command.extend(['-c:a', 'aac', '-b:a', '128k'])  # 音频编码为AAC
+        else:
+            logger.warning(f"视频 {input_path} 没有音频流，将会忽略音频设置")
+            command.extend(['-an'])  # 没有音频流时移除音频设置
+
+    # 视频处理参数：缩放并添加填充以保持比例
+    scale_filter = f"scale={target_width}:{target_height}:force_original_aspect_ratio=decrease"
+    pad_filter = f"pad={target_width}:{target_height}:(ow-iw)/2:(oh-ih)/2"
+    command.extend([
+        '-vf', f"{scale_filter},{pad_filter}",
+        '-r', '30',  # 设置帧率为30fps
+    ])
+
+    # 选择编码器
+    if hwaccel == 'cuda' or hwaccel == 'nvenc':
+        command.extend(['-c:v', 'h264_nvenc', '-preset', 'p4', '-profile:v', 'high'])
+    elif hwaccel == 'qsv':
+        command.extend(['-c:v', 'h264_qsv', '-preset', 'medium'])
+    elif hwaccel == 'videotoolbox':
+        command.extend(['-c:v', 'h264_videotoolbox', '-profile:v', 'high'])
+    elif hwaccel == 'vaapi':
+        command.extend(['-c:v', 'h264_vaapi', '-profile', '100'])
+    else:
+        command.extend(['-c:v', 'libx264', '-preset', 'medium', '-profile:v', 'high'])
+
+    # 设置视频比特率和其他参数
+    command.extend([
+        '-b:v', '5M',
+        '-maxrate', '8M',
+        '-bufsize', '10M',
+        '-pix_fmt', 'yuv420p',  # 兼容性更好的颜色格式
+    ])
+
+    # 输出文件
+    command.append(output_path)
+
+    # 执行命令
+    try:
+        logger.info(f"处理视频 {input_path} -> {output_path}")
+        subprocess.run(command, check=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
+        return output_path
+    except subprocess.CalledProcessError as e:
+        logger.error(f"处理视频失败: {e.stderr.decode() if e.stderr else str(e)}")
+        raise RuntimeError(f"处理视频失败: {str(e)}")
+
+
+def combine_clip_videos(
+        output_video_path: str,
+        video_paths: List[str],
+        video_ost_list: List[int],
+        video_aspect: VideoAspect = VideoAspect.portrait,
+        threads: int = 4,
+) -> str:
+    """
+    合并子视频
+    Args:
+        output_video_path: 合并后的存储路径
+        video_paths: 子视频路径列表
+        video_ost_list: 原声播放列表 (0: 不保留原声, 1: 只保留原声, 2: 保留原声并保留解说)
+        video_aspect: 屏幕比例
+        threads: 线程数
+
+    Returns:
+        str: 合并后的视频路径
+    """
+    # 检查ffmpeg是否安装
+    if not check_ffmpeg_installation():
+        raise RuntimeError("未找到ffmpeg，请先安装")
+
+    # 准备输出目录
+    output_dir = os.path.dirname(output_video_path)
+    os.makedirs(output_dir, exist_ok=True)
+
+    # 获取目标分辨率
+    aspect = VideoAspect(video_aspect)
+    video_width, video_height = aspect.to_resolution()
+
+    # 检测可用的硬件加速选项
+    hwaccel = get_hardware_acceleration_option()
+    if hwaccel:
+        logger.info(f"将使用 {hwaccel} 硬件加速")
+
+    # 重组视频路径和原声设置为一个字典列表结构
+    video_segments = []
+    
+    # 检查视频路径和原声设置列表长度是否匹配
+    if len(video_paths) != len(video_ost_list):
+        logger.warning(f"视频路径列表({len(video_paths)})和原声设置列表({len(video_ost_list)})长度不匹配")
+        # 调整长度以匹配较短的列表
+        min_length = min(len(video_paths), len(video_ost_list))
+        video_paths = video_paths[:min_length]
+        video_ost_list = video_ost_list[:min_length]
+    
+    # 创建视频处理配置字典列表
+    for i, (video_path, video_ost) in enumerate(zip(video_paths, video_ost_list)):
+        if not os.path.exists(video_path):
+            logger.warning(f"视频不存在，跳过: {video_path}")
+            continue
+            
+        # 检查是否有音频流
+        has_audio = check_video_has_audio(video_path)
+        
+        # 构建视频片段配置
+        segment = {
+            "index": i,
+            "path": video_path,
+            "ost": video_ost,
+            "has_audio": has_audio,
+            "keep_audio": video_ost > 0 and has_audio  # 只有当ost>0且实际有音频时才保留
+        }
+        
+        # 记录日志
+        if video_ost > 0 and not has_audio:
+            logger.warning(f"视频 {video_path} 设置为保留原声(ost={video_ost})，但该视频没有音频流")
+        
+        video_segments.append(segment)
+        
+    # 处理每个视频片段
+    processed_videos = []
+    temp_dir = os.path.join(output_dir, "temp_videos")
+    os.makedirs(temp_dir, exist_ok=True)
+
+    try:
+        # 第一阶段：处理所有视频片段到中间文件
+        for segment in video_segments:
+            # 处理单个视频，去除或保留音频
+            temp_output = os.path.join(temp_dir, f"processed_{segment['index']}.mp4")
+            try:
+                process_single_video(
+                    input_path=segment['path'],
+                    output_path=temp_output,
+                    target_width=video_width,
+                    target_height=video_height,
+                    keep_audio=segment['keep_audio'],
+                    hwaccel=hwaccel
+                )
+                processed_videos.append({
+                    "index": segment["index"],
+                    "path": temp_output,
+                    "keep_audio": segment["keep_audio"]
+                })
+                logger.info(f"视频 {segment['index'] + 1}/{len(video_segments)} 处理完成")
+            except Exception as e:
+                logger.error(f"处理视频 {segment['path']} 时出错: {str(e)}")
+                continue
+
+        if not processed_videos:
+            raise ValueError("没有有效的视频片段可以合并")
+            
+        # 按原始索引排序处理后的视频
+        processed_videos.sort(key=lambda x: x["index"])
+        
+        # 第二阶段：分步骤合并视频 - 避免复杂的filter_complex滤镜
+        try:
+            # 1. 首先，将所有没有音频的视频或音频被禁用的视频合并到一个临时文件中
+            video_paths_only = [video["path"] for video in processed_videos]
+            video_concat_path = os.path.join(temp_dir, "video_concat.mp4")
+            
+            # 创建concat文件，用于合并视频流
+            concat_file = os.path.join(temp_dir, "concat_list.txt")
+            create_ffmpeg_concat_file(video_paths_only, concat_file)
+            
+            # 合并所有视频流，但不包含音频
+            concat_cmd = [
+                'ffmpeg', '-y',
+                '-f', 'concat',
+                '-safe', '0',
+                '-i', concat_file,
+                '-c:v', 'libx264',
+                '-preset', 'medium',
+                '-profile:v', 'high',
+                '-an',  # 不包含音频
+                '-threads', str(threads),
+                video_concat_path
+            ]
+            
+            subprocess.run(concat_cmd, check=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
+            logger.info("视频流合并完成")
+            
+            # 2. 提取并合并有音频的片段
+            audio_segments = [video for video in processed_videos if video["keep_audio"]]
+            
+            if not audio_segments:
+                # 如果没有音频片段，直接使用无音频的合并视频作为最终结果
+                shutil.copy(video_concat_path, output_video_path)
+                logger.info("无音频视频合并完成")
+                return output_video_path
+            
+            # 创建音频中间文件
+            audio_files = []
+            for i, segment in enumerate(audio_segments):
+                # 提取音频
+                audio_file = os.path.join(temp_dir, f"audio_{i}.aac")
+                extract_audio_cmd = [
+                    'ffmpeg', '-y',
+                    '-i', segment["path"],
+                    '-vn',  # 不包含视频
+                    '-c:a', 'aac',
+                    '-b:a', '128k',
+                    audio_file
+                ]
+                subprocess.run(extract_audio_cmd, check=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
+                audio_files.append({
+                    "index": segment["index"],
+                    "path": audio_file
+                })
+                logger.info(f"提取音频 {i+1}/{len(audio_segments)} 完成")
+            
+            # 3. 计算每个音频片段的时间位置
+            audio_timings = []
+            current_time = 0.0
+            
+            # 获取每个视频片段的时长
+            for i, video in enumerate(processed_videos):
+                duration_cmd = [
+                    'ffprobe', '-v', 'error',
+                    '-show_entries', 'format=duration',
+                    '-of', 'csv=p=0',
+                    video["path"]
+                ]
+                result = subprocess.run(duration_cmd, check=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True)
+                duration = float(result.stdout.strip())
+                
+                # 如果当前片段需要保留音频，记录时间位置
+                if video["keep_audio"]:
+                    for audio in audio_files:
+                        if audio["index"] == video["index"]:
+                            audio_timings.append({
+                                "file": audio["path"],
+                                "start": current_time,
+                                "index": video["index"]
+                            })
+                            break
+                
+                current_time += duration
+            
+            # 4. 创建静音音频轨道作为基础
+            silence_audio = os.path.join(temp_dir, "silence.aac")
+            create_silence_cmd = [
+                'ffmpeg', '-y',
+                '-f', 'lavfi',
+                '-i', f'anullsrc=r=44100:cl=stereo',
+                '-t', str(current_time),  # 总时长
+                '-c:a', 'aac',
+                '-b:a', '128k',
+                silence_audio
+            ]
+            subprocess.run(create_silence_cmd, check=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
+            
+            # 5. 创建复杂滤镜命令以混合音频
+            filter_script = os.path.join(temp_dir, "filter_script.txt")
+            with open(filter_script, 'w') as f:
+                f.write(f"[0:a]volume=0.0[silence];\n")  # 首先静音背景轨道
+                
+                # 添加每个音频文件
+                for i, timing in enumerate(audio_timings):
+                    f.write(f"[{i+1}:a]adelay={int(timing['start']*1000)}|{int(timing['start']*1000)}[a{i}];\n")
+                
+                # 混合所有音频
+                mix_str = "[silence]"
+                for i in range(len(audio_timings)):
+                    mix_str += f"[a{i}]"
+                mix_str += f"amix=inputs={len(audio_timings)+1}:duration=longest[aout]"
+                f.write(mix_str)
+            
+            # 6. 构建音频合并命令
+            audio_inputs = ['-i', silence_audio]
+            for timing in audio_timings:
+                audio_inputs.extend(['-i', timing["file"]])
+                
+            mixed_audio = os.path.join(temp_dir, "mixed_audio.aac")
+            audio_mix_cmd = [
+                'ffmpeg', '-y'
+            ] + audio_inputs + [
+                '-filter_complex_script', filter_script,
+                '-map', '[aout]',
+                '-c:a', 'aac',
+                '-b:a', '128k',
+                mixed_audio
+            ]
+            
+            subprocess.run(audio_mix_cmd, check=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
+            logger.info("音频混合完成")
+            
+            # 7. 将合并的视频和混合的音频组合在一起
+            final_cmd = [
+                'ffmpeg', '-y',
+                '-i', video_concat_path,
+                '-i', mixed_audio,
+                '-c:v', 'copy',
+                '-c:a', 'aac',
+                '-map', '0:v:0',
+                '-map', '1:a:0',
+                '-shortest',
+                output_video_path
+            ]
+            
+            subprocess.run(final_cmd, check=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
+            logger.info("视频最终合并完成")
+            
+            return output_video_path
+            
+        except subprocess.CalledProcessError as e:
+            logger.error(f"合并视频过程中出错: {e.stderr.decode() if e.stderr else str(e)}")
+            
+            # 尝试备用合并方法 - 最简单的无音频合并
+            logger.info("尝试备用合并方法 - 无音频合并")
+            try:
+                concat_file = os.path.join(temp_dir, "concat_list.txt")
+                video_paths_only = [video["path"] for video in processed_videos]
+                create_ffmpeg_concat_file(video_paths_only, concat_file)
+                
+                backup_cmd = [
+                    'ffmpeg', '-y',
+                    '-f', 'concat',
+                    '-safe', '0',
+                    '-i', concat_file,
+                    '-c:v', 'copy',
+                    '-an',  # 无音频
+                    output_video_path
+                ]
+                
+                subprocess.run(backup_cmd, check=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
+                logger.warning("使用备用方法（无音频）成功合并视频")
+                return output_video_path
+            except Exception as backup_error:
+                logger.error(f"备用合并方法也失败: {str(backup_error)}")
+                raise RuntimeError(f"无法合并视频: {str(backup_error)}")
+            
+    except Exception as e:
+        logger.error(f"合并视频时出错: {str(e)}")
+        raise
+    finally:
+        # 清理临时文件
+        try:
+            if os.path.exists(temp_dir):
+                shutil.rmtree(temp_dir)
+                logger.info("已清理临时文件")
+        except Exception as e:
+            logger.warning(f"清理临时文件时出错: {str(e)}")
+
+
+if __name__ == '__main__':
+    video_paths = [
+        '/Users/apple/Desktop/home/NarratoAI/storage/temp/clip_video/0ac14d474144b54d614c26a5c87cffe7/vid-00-00-00-00-00-26.mp4',
+        '/Users/apple/Desktop/home/NarratoAI/storage/temp/clip_video/0ac14d474144b54d614c26a5c87cffe7/vid-00-01-15-00-01-29.mp4',
+        '/Users/apple/Desktop/home/NarratoAI/storage/temp/clip_video/6e7e343c7592c7d6f9a9636b55000f23/vid-00-04-41-00-04-58.mp4',
+        '/Users/apple/Desktop/home/NarratoAI/storage/temp/clip_video/0ac14d474144b54d614c26a5c87cffe7/vid-00-04-58-00-05-20.mp4',
+        '/Users/apple/Desktop/home/NarratoAI/storage/temp/clip_video/0ac14d474144b54d614c26a5c87cffe7/vid-00-05-45-00-05-53.mp4',
+        '/Users/apple/Desktop/home/NarratoAI/storage/temp/clip_video/6e7e343c7592c7d6f9a9636b55000f23/vid-00-06-00-00-06-03.mp4'
+    ]
+
+    combine_clip_videos(
+        output_video_path="/Users/apple/Desktop/home/NarratoAI/storage/temp/merge/merged_123.mp4",
+        video_paths=video_paths,
+        video_ost_list=[1, 0, 1, 0, 0, 1],
+        video_aspect=VideoAspect.portrait
+    )

app/services/subtitle_merger.py

@@ -0,0 +1,154 @@
+#!/usr/bin/env python
+# -*- coding: UTF-8 -*-
+
+'''
+@Project: NarratoAI
+@File   : subtitle_merger
+@Author : viccy
+@Date   : 2025/5/6 下午4:00 
+'''
+
+import re
+import os
+from datetime import datetime, timedelta
+
+
+def parse_time(time_str):
+    """解析时间字符串为timedelta对象"""
+    hours, minutes, seconds_ms = time_str.split(':')
+    seconds, milliseconds = seconds_ms.split(',')
+    
+    td = timedelta(
+        hours=int(hours),
+        minutes=int(minutes),
+        seconds=int(seconds),
+        milliseconds=int(milliseconds)
+    )
+    return td
+
+
+def format_time(td):
+    """将timedelta对象格式化为SRT时间字符串"""
+    total_seconds = int(td.total_seconds())
+    hours = total_seconds // 3600
+    minutes = (total_seconds % 3600) // 60
+    seconds = total_seconds % 60
+    milliseconds = td.microseconds // 1000
+    
+    return f"{hours:02d}:{minutes:02d}:{seconds:02d},{milliseconds:03d}"
+
+
+def extract_time_range_from_filename(filename):
+    """从文件名中提取时间范围"""
+    pattern = r'subtitle_(\d{2})_(\d{2})_(\d{2})-(\d{2})_(\d{2})_(\d{2})'
+    match = re.search(pattern, filename)
+    
+    if not match:
+        return None, None
+    
+    start_h, start_m, start_s, end_h, end_m, end_s = map(int, match.groups())
+    
+    start_time = timedelta(hours=start_h, minutes=start_m, seconds=start_s)
+    end_time = timedelta(hours=end_h, minutes=end_m, seconds=end_s)
+    
+    return start_time, end_time
+
+
+def merge_subtitle_files(subtitle_files, output_file=None):
+    """
+    合并多个SRT字幕文件
+    
+    参数:
+        subtitle_files: 包含SRT文件路径的列表
+        output_file: 输出文件的路径，如果为None则自动生成
+    
+    返回:
+        合并后的字幕文件路径
+    """
+    # 按文件名中的开始时间排序
+    sorted_files = sorted(subtitle_files, 
+                          key=lambda x: extract_time_range_from_filename(x)[0])
+    
+    merged_subtitles = []
+    subtitle_index = 1
+    
+    for file_path in sorted_files:
+        # 从文件名获取起始时间偏移
+        offset_time, _ = extract_time_range_from_filename(file_path)
+        
+        if offset_time is None:
+            print(f"警告: 无法从文件名 {os.path.basename(file_path)} 中提取时间范围，跳过该文件")
+            continue
+        
+        with open(file_path, 'r', encoding='utf-8') as file:
+            content = file.read()
+            
+        # 解析字幕文件
+        subtitle_blocks = re.split(r'\n\s*\n', content.strip())
+        
+        for block in subtitle_blocks:
+            lines = block.strip().split('\n')
+            if len(lines) < 3:  # 确保块有足够的行数
+                continue
+                
+            # 解析时间轴行
+            time_line = lines[1]
+            time_parts = time_line.split(' --> ')
+            if len(time_parts) != 2:
+                continue
+                
+            start_time = parse_time(time_parts[0])
+            end_time = parse_time(time_parts[1])
+            
+            # 应用时间偏移
+            adjusted_start_time = start_time + offset_time
+            adjusted_end_time = end_time + offset_time
+            
+            # 重建字幕块
+            adjusted_time_line = f"{format_time(adjusted_start_time)} --> {format_time(adjusted_end_time)}"
+            text_lines = lines[2:]
+            
+            new_block = [
+                str(subtitle_index),
+                adjusted_time_line,
+                *text_lines
+            ]
+            
+            merged_subtitles.append('\n'.join(new_block))
+            subtitle_index += 1
+    
+    # 合并所有字幕块
+    merged_content = '\n\n'.join(merged_subtitles)
+    
+    # 确定输出文件路径
+    if output_file is None:
+        # 自动生成输出文件名
+        first_file_path = sorted_files[0]
+        last_file_path = sorted_files[-1]
+        _, first_end = extract_time_range_from_filename(first_file_path)
+        _, last_end = extract_time_range_from_filename(last_file_path)
+        
+        dir_path = os.path.dirname(first_file_path)
+        first_start_str = os.path.basename(first_file_path).split('-')[0].replace('subtitle_', '')
+        last_end_h, last_end_m, last_end_s = int(last_end.seconds // 3600), int((last_end.seconds % 3600) // 60), int(last_end.seconds % 60)
+        last_end_str = f"{last_end_h:02d}_{last_end_m:02d}_{last_end_s:02d}"
+        
+        output_file = os.path.join(dir_path, f"merged_subtitle_{first_start_str}-{last_end_str}.srt")
+    
+    # 写入合并后的内容
+    with open(output_file, 'w', encoding='utf-8') as file:
+        file.write(merged_content)
+    
+    return output_file
+
+
+if __name__ == '__main__':
+    subtitle_files = [
+        "/Users/apple/Desktop/home/NarratoAI/storage/tasks/qyn2-2-demo/subtitle_00_00_00-00_01_15.srt",
+        "/Users/apple/Desktop/home/NarratoAI/storage/tasks/qyn2-2-demo/subtitle_00_01_15-00_04_40.srt",
+        "/Users/apple/Desktop/home/NarratoAI/storage/tasks/qyn2-2-demo/subtitle_00_04_58-00_05_45.srt",
+        "/Users/apple/Desktop/home/NarratoAI/storage/tasks/qyn2-2-demo/subtitle_00_05_45-00_06_00.srt",
+    ]
+    
+    output_file = merge_subtitle_files(subtitle_files)
+    print(f"字幕文件已合并至: {output_file}")

app/services/task.py

@@ -9,7 +9,7 @@ from loguru import logger
 from app.config import config
 from app.models import const
 from app.models.schema import VideoConcatMode, VideoParams, VideoClipParams
-from app.services import llm, material, subtitle, video, voice, audio_merger
+from app.services import llm, material, subtitle, video, voice, audio_merger, subtitle_merger, clip_video
 from app.services import state as sm
 from app.utils import utils
 
@@ -158,18 +158,25 @@ def get_video_materials(task_id, params, video_terms, audio_duration):
 
 
 def start_subclip(task_id: str, params: VideoClipParams, subclip_path_videos: dict):
-    """后台任务（自动剪辑视频进行剪辑）"""
+    """
+    后台任务（自动剪辑视频进行剪辑）
+    Args:
+        task_id: 任务ID
+        params: 视频参数
+        subclip_path_videos: 视频片段路径
+    """
     logger.info(f"\n\n## 开始任务: {task_id}")
-    
+    sm.state.update_task(task_id, state=const.TASK_STATE_PROCESSING, progress=0)
-    # 初始化 ImageMagick
-    if not utils.init_imagemagick():
-        logger.warning("ImageMagick 初始化失败，字幕可能无法正常显示")
-    
-    sm.state.update_task(task_id, state=const.TASK_STATE_PROCESSING, progress=5)
 
-    # tts 角色名称
+    # # 初始化 ImageMagick
-    voice_name = voice.parse_voice_name(params.voice_name)
+    # if not utils.init_imagemagick():
+    #     logger.warning("ImageMagick 初始化失败，字幕可能无法正常显示")
 
+    # # tts 角色名称
+    # voice_name = voice.parse_voice_name(params.voice_name)
+    """
+    1. 加载剪辑脚本
+    """
     logger.info("\n\n## 1. 加载视频脚本")
     video_script_path = path.join(params.video_clip_json_path)
     
@@ -187,111 +194,102 @@ def start_subclip(task_id: str, params: VideoClipParams, subclip_path_videos: di
                 logger.debug(f"解说时间戳列表: \n{time_list}")
                 
                 # 获取视频总时长(单位 s)
-                last_timestamp = list_script[-1]['new_timestamp']
+                last_timestamp = list_script[-1]['timestamp'].split("-")[1]
-                end_time = last_timestamp.split("-")[1]
+                total_duration = utils.time_to_seconds(last_timestamp)
-                total_duration = utils.time_to_seconds(end_time)
+
-                
         except Exception as e:
-            logger.error(f"无法读取视频json脚本，请检查配置是否正确。{e}")
+            logger.error(f"无法读取视频json脚本，请检查脚本格式是否正确")
-            raise ValueError("无法读取视频json脚本，请检查配置是否正确")
+            raise ValueError("无法读取视频json脚本，请检查脚本格式是否正确")
     else:
         logger.error(f"video_script_path: {video_script_path} \n\n", traceback.format_exc())
         raise ValueError("解说脚本不存在！请检查配置是否正确。")
 
+    """
+    2. 使用 TTS 生成音频素材
+    """
     logger.info("\n\n## 2. 根据OST设置生成音频列表")
-    # 只为OST=0或2的片段生成TTS音频
+    # 只为OST=0 or 2的判断生成音频， OST=0 仅保留解说 OST=2 保留解说和原声
     tts_segments = [
         segment for segment in list_script 
         if segment['OST'] in [0, 2]
     ]
     logger.debug(f"需要生成TTS的片段数: {len(tts_segments)}")
-    
-    # 初始化音频文件路径
-    audio_files = []
-    final_audio = ""
-    
-    if tts_segments:
-        audio_files, sub_maker_list = voice.tts_multiple(
-            task_id=task_id,
-            list_script=tts_segments,  # 只传入需要TTS的片段
-            voice_name=voice_name,
-            voice_rate=params.voice_rate,
-            voice_pitch=params.voice_pitch,
-            force_regenerate=True
-        )
-        if audio_files is None:
-            sm.state.update_task(task_id, state=const.TASK_STATE_FAILED)
-            logger.error("TTS转换音频失败, 可能是网络不可用! 如果您在中国, 请使用VPN.")
-            return
 
-        if audio_files:
+    tts_results = voice.tts_multiple(
-            logger.info(f"合并音频文件: {audio_files}")
+        task_id=task_id,
-            try:
+        list_script=tts_segments,  # 只传入需要TTS的片段
-                # 传入OST信息以便正确处理音频
+        voice_name=params.voice_name,
-                final_audio = audio_merger.merge_audio_files(
+        voice_rate=params.voice_rate,
-                    task_id=task_id,
+        voice_pitch=params.voice_pitch,
-                    audio_files=audio_files,
+        force_regenerate=True
-                    total_duration=total_duration,
+    )
-                    list_script=list_script  # 传入完整脚本以便处理OST
+    audio_files = [
-                )
+        tts_result["audio_file"] for tts_result in tts_results
-                logger.info("音频文件合并成功")
+    ]
-            except Exception as e:
+    subtitle_files = [
-                logger.error(f"合并音频文件失败: {str(e)}")
+        tts_result["subtitle_file"] for tts_result in tts_results
-                final_audio = ""
+    ]
-    else:
+    if tts_results:
-        # 如果没有需要生成TTS的片段，创建一个空白音频文件
+        logger.info(f"合并音频/字幕文件")
-        # 这样可以确保后续的音频处理能正确进行
-        logger.info("没有需要生成TTS的片段，将保留原声和背景音乐")
-        final_audio = path.join(utils.task_dir(task_id), "empty.mp3")
         try:
-            from moviepy.editor import AudioClip
+            # 合并音频文件
-            # 创建一个与视频等长的空白音频
+            merged_audio_path = audio_merger.merge_audio_files(
-            empty_audio = AudioClip(make_frame=lambda t: 0, duration=total_duration)
+                task_id=task_id,
-            empty_audio.write_audiofile(final_audio, fps=44100)
+                audio_files=audio_files,
-            logger.info(f"已创建空白音频文件: {final_audio}")
+                total_duration=total_duration,
-        except Exception as e:
+                list_script=list_script  # 传入完整脚本以便处理OST
-            logger.error(f"创建空白音频文件失败: {str(e)}")
-            final_audio = ""
-
-    sm.state.update_task(task_id, state=const.TASK_STATE_PROCESSING, progress=30)
-
-    subtitle_path = ""
-    if params.subtitle_enabled:
-        if audio_files:
-            subtitle_path = path.join(utils.task_dir(task_id), f"subtitle.srt")
-            subtitle_provider = config.app.get("subtitle_provider", "").strip().lower()
-            logger.info(f"\n\n## 3. 生成字幕、提供程序是: {subtitle_provider}")
-
-            subtitle.create(
-                audio_file=final_audio,
-                subtitle_file=subtitle_path,
             )
+            logger.info(f"音频文件合并成功->{merged_audio_path}")
+            # 合并字幕文件
+            merged_subtitle_path = subtitle_merger.merge_subtitle_files(
+                subtitle_files=subtitle_files,
+            )
+            logger.info(f"字幕文件合并成功->{merged_subtitle_path}")
+        except Exception as e:
+            logger.error(f"合并音频文件失败: {str(e)}")
+            merged_audio_path = ""
+            merged_subtitle_path = ""
+    else:
+        logger.error("TTS转换音频失败, 可能是网络不可用! 如果您在中国, 请使用VPN.")
+        return
+    sm.state.update_task(task_id, state=const.TASK_STATE_PROCESSING, progress=20)
 
-            subtitle_lines = subtitle.file_to_subtitles(subtitle_path)
+    """
+    3. (可选) 使用 whisper 生成字幕
+    """
+    if merged_subtitle_path is None:
+        if audio_files:
+            merged_subtitle_path = path.join(utils.task_dir(task_id), f"subtitle.srt")
+            subtitle_provider = config.app.get("subtitle_provider", "").strip().lower()
+            logger.info(f"\n\n使用 {subtitle_provider} 生成字幕")
+            
+            subtitle.create(
+                audio_file=merged_audio_path,
+                subtitle_file=merged_subtitle_path,
+            )
+            subtitle_lines = subtitle.file_to_subtitles(merged_subtitle_path)
             if not subtitle_lines:
-                logger.warning(f"字幕文件无效: {subtitle_path}")
+                logger.warning(f"字幕文件无效: {merged_subtitle_path}")
-                subtitle_path = ""
 
     sm.state.update_task(task_id, state=const.TASK_STATE_PROCESSING, progress=40)
 
+    """
+    4. 裁剪视频 - 将超出音频长度的视频进行裁剪
+    """
     logger.info("\n\n## 4. 裁剪视频")
+    result = clip_video.clip_video(params.video_origin_path, tts_results)
+    subclip_path_videos.update(result)
     subclip_videos = [x for x in subclip_path_videos.values()]
-    # logger.debug(f"\n\n## 裁剪后的视频文件列表: \n{subclip_videos}")
 
-    if not subclip_videos:
+    sm.state.update_task(task_id, state=const.TASK_STATE_PROCESSING, progress=60)
-        sm.state.update_task(task_id, state=const.TASK_STATE_FAILED)
-        logger.error(
-            "裁剪视频失败，可能是 ImageMagick 不可用")
-        return
-
-    sm.state.update_task(task_id, state=const.TASK_STATE_PROCESSING, progress=50)
 
+    """
+    5. 合并视频
+    """
     final_video_paths = []
     combined_video_paths = []
 
-    _progress = 50
+    combined_video_path = path.join(utils.task_dir(task_id), f"merger.mp4")
-    index = 1
-    combined_video_path = path.join(utils.task_dir(task_id), f"combined.mp4")
     logger.info(f"\n\n## 5. 合并视频: => {combined_video_path}")
 
     video.combine_clip_videos(
@@ -302,14 +300,15 @@ def start_subclip(task_id: str, params: VideoClipParams, subclip_path_videos: di
         video_aspect=params.video_aspect,
         threads=params.n_threads  # 多线程
     )
+    sm.state.update_task(task_id, state=const.TASK_STATE_PROCESSING, progress=80)
 
-    _progress += 50 / 2
-    sm.state.update_task(task_id, progress=_progress)
 
-    final_video_path = path.join(utils.task_dir(task_id), f"final-{index}.mp4")
+    """
+    6. 合并字幕/BGM/配音/视频
+    """
+    final_video_path = path.join(utils.task_dir(task_id), f"combined.mp4")
+    logger.info(f"\n\n## 6. 最后一步: 合并字幕/BGM/配音/视频 -> {final_video_path}")
 
-    logger.info(f"\n\n## 6. 最后合成: {index} => {final_video_path}")
-    
     # 获取背景音乐
     bgm_path = None
     if params.bgm_type or params.bgm_file:
@@ -340,18 +339,15 @@ def start_subclip(task_id: str, params: VideoClipParams, subclip_path_videos: di
     font_path = utils.font_dir(params.font_name)
     video.generate_video_v3(
         video_path=combined_video_path,
-        subtitle_path=subtitle_path,
+        subtitle_path=merged_subtitle_path,
         bgm_path=bgm_path,
-        narration_path=final_audio,
+        narration_path=merged_audio_path,
         output_path=final_video_path,
         volume_config=volume_config,  # 添加音量配置
         subtitle_style=subtitle_style,
         font_path=font_path
     )
 
-    _progress += 50 / 2
-    sm.state.update_task(task_id, progress=_progress)
-
     final_video_paths.append(final_video_path)
     combined_video_paths.append(combined_video_path)
 
@@ -400,35 +396,20 @@ def validate_params(video_path, audio_path, output_file, params):
 
 
 if __name__ == "__main__":
-    # task_id = "test123"
+    task_id = "qyn2-2-demo"
-    # subclip_path_videos = {'00:41-01:58': 'E:\\projects\\NarratoAI\\storage\\cache_videos/vid-00_41-01_58.mp4',
-    #                        '00:06-00:15': 'E:\\projects\\NarratoAI\\storage\\cache_videos/vid-00_06-00_15.mp4',
-    #                        '01:10-01:17': 'E:\\projects\\NarratoAI\\storage\\cache_videos/vid-01_10-01_17.mp4',
-    #                        '00:47-01:03': 'E:\\projects\\NarratoAI\\storage\\cache_videos/vid-00_47-01_03.mp4',
-    #                        '01:03-01:10': 'E:\\projects\\NarratoAI\\storage\\cache_videos/vid-01_03-01_10.mp4',
-    #                        '02:40-03:08': 'E:\\projects\\NarratoAI\\storage\\cache_videos/vid-02_40-03_08.mp4',
-    #                        '03:02-03:20': 'E:\\projects\\NarratoAI\\storage\\cache_videos/vid-03_02-03_20.mp4',
-    #                        '03:18-03:20': 'E:\\projects\\NarratoAI\\storage\\cache_videos/vid-03_18-03_20.mp4'}
-    #
-    # params = VideoClipParams(
-    #     video_clip_json_path="E:\\projects\\NarratoAI\\resource/scripts/test003.json",
-    #     video_origin_path="E:\\projects\\NarratoAI\\resource/videos/1.mp4",
-    # )
-    # start_subclip(task_id, params, subclip_path_videos=subclip_path_videos)
 
-    task_id = "test456"
+    # 提前裁剪是为了方便检查视频
-    subclip_path_videos = {'01:10-01:17': './storage/cache_videos/vid-01_10-01_17.mp4',
+    subclip_path_videos = {
-                           '01:58-02:04': './storage/cache_videos/vid-01_58-02_04.mp4',
+        '00:00:00-00:01:15': '/Users/apple/Desktop/home/NarratoAI/storage/temp/clip_video/6e7e343c7592c7d6f9a9636b55000f23/vid-00-00-00-00-01-15.mp4',
-                           '02:25-02:31': './storage/cache_videos/vid-02_25-02_31.mp4',
+        '00:01:15-00:04:40': '/Users/apple/Desktop/home/NarratoAI/storage/temp/clip_video/6e7e343c7592c7d6f9a9636b55000f23/vid-00-01-15-00-04-40.mp4',
-                           '01:28-01:33': './storage/cache_videos/vid-01_28-01_33.mp4',
+        '00:04:41-00:04:58': '/Users/apple/Desktop/home/NarratoAI/storage/temp/clip_video/6e7e343c7592c7d6f9a9636b55000f23/vid-00-04-41-00-04-58.mp4',
-                           '03:14-03:18': './storage/cache_videos/vid-03_14-03_18.mp4',
+        '00:04:58-00:05:45': '/Users/apple/Desktop/home/NarratoAI/storage/temp/clip_video/6e7e343c7592c7d6f9a9636b55000f23/vid-00-04-58-00-05-45.mp4',
-                           '00:24-00:28': './storage/cache_videos/vid-00_24-00_28.mp4',
+        '00:05:45-00:06:00': '/Users/apple/Desktop/home/NarratoAI/storage/temp/clip_video/6e7e343c7592c7d6f9a9636b55000f23/vid-00-05-45-00-06-00.mp4',
-                           '03:02-03:08': './storage/cache_videos/vid-03_02-03_08.mp4',
+        '00:06:00-00:06:03': '/Users/apple/Desktop/home/NarratoAI/storage/temp/clip_video/6e7e343c7592c7d6f9a9636b55000f23/vid-00-06-00-00-06-03.mp4',
-                           '00:41-00:44': './storage/cache_videos/vid-00_41-00_44.mp4',
+    }
-                           '02:12-02:25': './storage/cache_videos/vid-02_12-02_25.mp4'}
 
     params = VideoClipParams(
-        video_clip_json_path="/Users/apple/Desktop/home/NarratoAI/resource/scripts/test004.json",
+        video_clip_json_path="/Users/apple/Desktop/home/NarratoAI/resource/scripts/demo.json",
-        video_origin_path="/Users/apple/Desktop/home/NarratoAI/resource/videos/1.mp4",
+        video_origin_path="/Users/apple/Desktop/home/NarratoAI/resource/videos/qyn2-2无片头片尾.mp4",
     )
-    start_subclip(task_id, params, subclip_path_videos=subclip_path_videos)
+    start_subclip(task_id, params, subclip_path_videos)

app/services/video.py

@@ -443,4 +443,3 @@ def generate_video_v3(
         bgm.close()
     if narration_path:
         narration.close()
-

app/services/voice.py

@@ -5,10 +5,11 @@ import traceback
 import edge_tts
 import asyncio
 from loguru import logger
-from typing import List
+from typing import List, Union
 from datetime import datetime
 from xml.sax.saxutils import unescape
 from edge_tts import submaker, SubMaker
+from edge_tts.submaker import mktimestamp
 from moviepy.video.tools import subtitles
 import time
 
@@ -1036,7 +1037,7 @@ def is_azure_v2_voice(voice_name: str):
 
 def tts(
     text: str, voice_name: str, voice_rate: float, voice_pitch: float, voice_file: str
-) -> [SubMaker, None]:
+) -> Union[SubMaker, None]:
     if is_azure_v2_voice(voice_name):
         return azure_tts_v2(text, voice_name, voice_file)
     return azure_tts_v1(text, voice_name, voice_rate, voice_pitch, voice_file)
@@ -1064,7 +1065,7 @@ def convert_pitch_to_percent(rate: float) -> str:
 
 def azure_tts_v1(
     text: str, voice_name: str, voice_rate: float, voice_pitch: float, voice_file: str
-) -> [SubMaker, None]:
+) -> Union[SubMaker, None]:
     voice_name = parse_voice_name(voice_name)
     text = text.strip()
     rate_str = convert_rate_to_percent(voice_rate)
@@ -1087,11 +1088,6 @@ def azure_tts_v1(
                         )
                 return sub_maker, audio_data
 
-            # 判断音频文件是否已存在
-            if os.path.exists(voice_file):
-                logger.info(f"voice file exists, skip tts: {voice_file}")
-                continue
-
             # 获取音频数据和字幕信息
             sub_maker, audio_data = asyncio.run(_do())
             
@@ -1105,8 +1101,6 @@ def azure_tts_v1(
             # 数据有效，写入文件
             with open(voice_file, "wb") as file:
                 file.write(audio_data)
-
-            logger.info(f"completed, output file: {voice_file}")
             return sub_maker
         except Exception as e:
             logger.error(f"生成音频文件时出错: {str(e)}")
@@ -1115,7 +1109,7 @@ def azure_tts_v1(
     return None
 
 
-def azure_tts_v2(text: str, voice_name: str, voice_file: str) -> [SubMaker, None]:
+def azure_tts_v2(text: str, voice_name: str, voice_file: str) -> Union[SubMaker, None]:
     voice_name = is_azure_v2_voice(voice_name)
     if not voice_name:
         logger.error(f"invalid voice name: {voice_name}")
@@ -1240,7 +1234,7 @@ def create_subtitle_from_multiple(text: str, sub_maker_list: List[SubMaker], lis
             if script_item['OST']:
                 continue
 
-            start_time, end_time = script_item['new_timestamp'].split('-')
+            start_time, end_time = script_item['timestamp'].split('-')
             if sub_maker_index >= len(sub_maker_list):
                 logger.error(f"Sub maker list index out of range: {sub_maker_index}")
                 break
@@ -1317,6 +1311,97 @@ def create_subtitle_from_multiple(text: str, sub_maker_list: List[SubMaker], lis
         traceback.print_exc()
 
 
+def create_subtitle(sub_maker: submaker.SubMaker, text: str, subtitle_file: str):
+    """
+    优化字幕文件
+    1. 将字幕文件按照标点符号分割成多行
+    2. 逐行匹配字幕文件中的文本
+    3. 生成新的字幕文件
+    """
+
+    text = _format_text(text)
+
+    def formatter(idx: int, start_time: float, end_time: float, sub_text: str) -> str:
+        """
+        1
+        00:00:00,000 --> 00:00:02,360
+        跑步是一项简单易行的运动
+        """
+        start_t = mktimestamp(start_time).replace(".", ",")
+        end_t = mktimestamp(end_time).replace(".", ",")
+        return f"{idx}\n" f"{start_t} --> {end_t}\n" f"{sub_text}\n"
+
+    start_time = -1.0
+    sub_items = []
+    sub_index = 0
+
+    script_lines = utils.split_string_by_punctuations(text)
+
+    def match_line(_sub_line: str, _sub_index: int):
+        if len(script_lines) <= _sub_index:
+            return ""
+
+        _line = script_lines[_sub_index]
+        if _sub_line == _line:
+            return script_lines[_sub_index].strip()
+
+        _sub_line_ = re.sub(r"[^\w\s]", "", _sub_line)
+        _line_ = re.sub(r"[^\w\s]", "", _line)
+        if _sub_line_ == _line_:
+            return _line_.strip()
+
+        _sub_line_ = re.sub(r"\W+", "", _sub_line)
+        _line_ = re.sub(r"\W+", "", _line)
+        if _sub_line_ == _line_:
+            return _line.strip()
+
+        return ""
+
+    sub_line = ""
+
+    try:
+        for _, (offset, sub) in enumerate(zip(sub_maker.offset, sub_maker.subs)):
+            _start_time, end_time = offset
+            if start_time < 0:
+                start_time = _start_time
+
+            sub = unescape(sub)
+            sub_line += sub
+            sub_text = match_line(sub_line, sub_index)
+            if sub_text:
+                sub_index += 1
+                line = formatter(
+                    idx=sub_index,
+                    start_time=start_time,
+                    end_time=end_time,
+                    sub_text=sub_text,
+                )
+                sub_items.append(line)
+                start_time = -1.0
+                sub_line = ""
+
+        if len(sub_items) == len(script_lines):
+            with open(subtitle_file, "w", encoding="utf-8") as file:
+                file.write("\n".join(sub_items) + "\n")
+            try:
+                sbs = subtitles.file_to_subtitles(subtitle_file, encoding="utf-8")
+                duration = max([tb for ((ta, tb), txt) in sbs])
+                logger.info(
+                    f"已创建字幕文件: {subtitle_file}, duration: {duration}"
+                )
+                return subtitle_file, duration
+            except Exception as e:
+                logger.error(f"failed, error: {str(e)}")
+                os.remove(subtitle_file)
+        else:
+            logger.warning(
+                f"字幕创建失败, 字幕长度: {len(sub_items)}, script_lines len: {len(script_lines)}"
+            )
+
+    except Exception as e:
+        logger.error(f"failed, error: {str(e)}")
+
+
 def get_audio_duration(sub_maker: submaker.SubMaker):
     """
     获取音频时长
@@ -1339,20 +1424,25 @@ def tts_multiple(task_id: str, list_script: list, voice_name: str, voice_rate: f
     """
     voice_name = parse_voice_name(voice_name)
     output_dir = utils.task_dir(task_id)
-    audio_files = []
+    tts_results = []
-    sub_maker_list = []
 
     for item in list_script:
+        tts_item = {
+            "audio_file": "",
+            "subtitle_file": "",
+            "duration": 0,
+        }
         if item['OST'] != 1:
             # 将时间戳中的冒号替换为下划线
-            timestamp = item['new_timestamp'].replace(':', '_')
+            timestamp = item['timestamp'].replace(':', '_')
             audio_file = os.path.join(output_dir, f"audio_{timestamp}.mp3")
+            subtitle_file = os.path.join(output_dir, f"subtitle_{timestamp}.srt")
             
-            # 检查文件是否已存在，如存在且不强制重新生成，则跳过
+            # # 检查文件是否已存在，如存在且不强制重新生成，则跳过
-            if os.path.exists(audio_file) and not force_regenerate:
+            # if os.path.exists(audio_file) and not force_regenerate:
-                logger.info(f"音频文件已存在，跳过生成: {audio_file}")
+            #     logger.info(f"音频文件已存在，跳过生成: {audio_file}")
-                audio_files.append(audio_file)
+            #     tts_item["audio_file"] = audio_file
-                continue
+            #     continue
 
             text = item['narration']
 
@@ -1369,9 +1459,17 @@ def tts_multiple(task_id: str, list_script: list, voice_name: str, voice_rate: f
                              f"如果您在中国，请使用VPN; "
                              f"或者使用其他 tts 引擎")
                 continue
+            else:
+                # 为当前片段生成字幕文件
+                _, duration = create_subtitle(sub_maker=sub_maker, text=text, subtitle_file=subtitle_file)
 
-            audio_files.append(audio_file)
+            tts_results.append({
-            sub_maker_list.append(sub_maker)
+                "timestamp": item['timestamp'],
+                "audio_file": audio_file,
+                "subtitle_file": subtitle_file,
+                "duration": duration,
+                "text": text,
+            })
             logger.info(f"已生成音频文件: {audio_file}")
 
-    return audio_files, sub_maker_list
+    return tts_results

webui.py

@@ -3,7 +3,8 @@ import os
 import sys
 from uuid import uuid4
 from app.config import config
-from webui.components import basic_settings, video_settings, audio_settings, subtitle_settings, script_settings, review_settings, merge_settings, system_settings
+from webui.components import basic_settings, video_settings, audio_settings, subtitle_settings, script_settings, \
+    review_settings, merge_settings, system_settings
 from webui.utils import cache, file_utils
 from app.utils import utils
 from app.models.schema import VideoClipParams, VideoAspect
@@ -28,6 +29,7 @@ hide_streamlit_style = """
 """
 st.markdown(hide_streamlit_style, unsafe_allow_html=True)
 
+
 def init_log():
     """初始化日志配置"""
     from loguru import logger
@@ -41,11 +43,11 @@ def init_log():
             "torch.cuda.is_available()",
             "CUDA initialization"
         ]
-        
+
         for msg in ignore_messages:
             if msg in record["message"]:
                 return ""
-            
+
         file_path = record["file"].path
         relative_path = os.path.relpath(file_path, config.root_dir)
         record["file"].path = f"./{relative_path}"
@@ -74,6 +76,7 @@ def init_log():
         filter=log_filter
     )
 
+
 def init_global_state():
     """初始化全局状态"""
     if 'video_clip_json' not in st.session_state:
@@ -85,6 +88,7 @@ def init_global_state():
     if 'subclip_videos' not in st.session_state:
         st.session_state['subclip_videos'] = {}
 
+
 def tr(key):
     """翻译函数"""
     i18n_dir = os.path.join(os.path.dirname(__file__), "webui", "i18n")
@@ -92,13 +96,14 @@ def tr(key):
     loc = locales.get(st.session_state['ui_language'], {})
     return loc.get("Translation", {}).get(key, key)
 
+
 def render_generate_button():
     """渲染生成按钮和处理逻辑"""
     if st.button(tr("Generate Video"), use_container_width=True, type="primary"):
         try:
             from app.services import task as tm
             import torch
-            
+
             # 重置日志容器和记录
             log_container = st.empty()
             log_records = []
@@ -152,7 +157,7 @@ def render_generate_button():
 
             video_files = result.get("videos", [])
             st.success(tr("视生成完成"))
-            
+
             try:
                 if video_files:
                     player_cols = st.columns(len(video_files) * 2 + 1)
@@ -167,15 +172,16 @@ def render_generate_button():
         finally:
             PerformanceMonitor.cleanup_resources()
 
+
 def main():
     """主函数"""
     init_log()
     init_global_state()
     utils.init_resources()
-    
+
     st.title(f"NarratoAI :sunglasses:📽️")
     st.write(tr("Get Help"))
-    
+
     # 渲染基础设置面板
     basic_settings.render_basic_settings(tr)
     # 渲染合并设置
@@ -192,12 +198,13 @@ def main():
         subtitle_settings.render_subtitle_panel(tr)
         # 渲染系统设置面板
         system_settings.render_system_panel(tr)
-    
+
     # 渲染视频审查面板
     review_settings.render_review_panel(tr)
-    
+
     # 渲染生成按钮和处理逻辑
     render_generate_button()
 
+
 if __name__ == "__main__":
     main()