feat: 新增脚本自动字幕生成功能并优化网页视频展示

添加script_subtitle服务，支持基于脚本内容自动生成标准SRT字幕文件
修改任务处理流程，优先使用新的脚本字幕生成逻辑，失败时回退至原TTS字幕合并方案
优化最终视频自动转录逻辑，已生成脚本字幕时跳过重复的自动转录步骤
改进网页端弹窗视频的展示样式，根据宽高比调整预览宽度并添加黑色背景
新增完整的单元测试覆盖字幕生成相关功能

app/services/script_subtitle.py

@@ -0,0 +1,213 @@
+import os
+import re
+import unicodedata
+from typing import Iterable, List, Optional, Sequence, Tuple
+
+from loguru import logger
+
+from app.utils import utils
+
+
+DEFAULT_SUBTITLE_OST_TYPES = (0, 2)
+DEFAULT_MAX_CHARS_PER_SUBTITLE = 12
+SENTENCE_PART_RE = re.compile(r"[^。！？!?；;，,、\n]+[。！？!?；;，,、]?")
+
+
+def _normalize_text(text: str) -> str:
+    return re.sub(r"\s+", " ", str(text or "")).strip()
+
+
+def _remove_punctuation(text: str) -> str:
+    return "".join(
+        char for char in str(text or "")
+        if not unicodedata.category(char).startswith("P")
+    )
+
+
+def clean_subtitle_text(text: str) -> str:
+    """Normalize subtitle text for burn-in display."""
+    return _normalize_text(_remove_punctuation(text))
+
+
+def split_narration(text: str, max_chars: int = DEFAULT_MAX_CHARS_PER_SUBTITLE) -> List[str]:
+    """Split narration into readable subtitle chunks."""
+    text = _normalize_text(text)
+    if not text:
+        return []
+
+    max_chars = max(1, int(max_chars or DEFAULT_MAX_CHARS_PER_SUBTITLE))
+    parts = [match.group(0).strip() for match in SENTENCE_PART_RE.finditer(text)]
+    if not parts:
+        parts = [text]
+
+    chunks = []
+    current = ""
+
+    def flush_long_part(part: str) -> str:
+        while len(part) > max_chars:
+            chunks.append(part[:max_chars].strip())
+            part = part[max_chars:].strip()
+        return part
+
+    for part in parts:
+        if not part:
+            continue
+
+        if len(part) > max_chars:
+            if current:
+                chunks.append(current.strip())
+                current = ""
+            current = flush_long_part(part)
+            continue
+
+        candidate = f"{current}{part}" if current else part
+        if len(candidate) <= max_chars:
+            current = candidate
+        else:
+            if current:
+                chunks.append(current.strip())
+            current = part
+
+    if current:
+        chunks.append(current.strip())
+
+    return [cleaned for chunk in chunks if (cleaned := clean_subtitle_text(chunk))]
+
+
+def parse_srt_like_time(time_text: str) -> float:
+    time_text = str(time_text or "").strip().replace(",", ".")
+    parts = time_text.split(":")
+    if len(parts) != 3:
+        raise ValueError(f"不支持的时间格式: {time_text}")
+
+    hours = int(parts[0])
+    minutes = int(parts[1])
+    seconds = float(parts[2])
+    return hours * 3600 + minutes * 60 + seconds
+
+
+def parse_time_range(time_range: str) -> Tuple[float, float]:
+    if not time_range or "-" not in str(time_range):
+        raise ValueError(f"不支持的时间范围: {time_range}")
+
+    start_text, end_text = str(time_range).split("-", 1)
+    start = parse_srt_like_time(start_text)
+    end = parse_srt_like_time(end_text)
+    if end <= start:
+        raise ValueError(f"结束时间必须晚于开始时间: {time_range}")
+
+    return start, end
+
+
+def format_srt_time(seconds: float) -> str:
+    milliseconds_total = max(0, int(round(float(seconds) * 1000)))
+    milliseconds = milliseconds_total % 1000
+    total_seconds = milliseconds_total // 1000
+    hours = total_seconds // 3600
+    minutes = (total_seconds % 3600) // 60
+    secs = total_seconds % 60
+    return f"{hours:02d}:{minutes:02d}:{secs:02d},{milliseconds:03d}"
+
+
+def _safe_ost_value(value) -> Optional[int]:
+    try:
+        return int(value)
+    except (TypeError, ValueError):
+        return None
+
+
+def _resolve_item_time_range(item: dict, current_time: float) -> Tuple[Optional[Tuple[float, float]], float]:
+    edited_time_range = item.get("editedTimeRange")
+    if edited_time_range:
+        try:
+            start, end = parse_time_range(edited_time_range)
+            return (start, end), end
+        except ValueError as e:
+            logger.warning(f"解析 editedTimeRange 失败，将尝试使用 duration: {e}")
+
+    duration = float(item.get("duration", 0.0) or 0.0)
+    if duration <= 0:
+        return None, current_time
+
+    start = current_time
+    end = current_time + duration
+    return (start, end), end
+
+
+def _build_srt_blocks(
+    list_script: Sequence[dict],
+    include_ost: Iterable[int],
+    max_chars: int,
+) -> List[str]:
+    include_ost_set = {int(item) for item in include_ost}
+    blocks = []
+    subtitle_index = 1
+    current_time = 0.0
+
+    for item in list_script:
+        time_range, current_time = _resolve_item_time_range(item, current_time)
+        if not time_range:
+            continue
+
+        ost = _safe_ost_value(item.get("OST"))
+        if ost not in include_ost_set:
+            continue
+
+        chunks = split_narration(item.get("narration", ""), max_chars=max_chars)
+        if not chunks:
+            continue
+
+        start, end = time_range
+        segment_duration = end - start
+        if segment_duration <= 0:
+            continue
+
+        chunk_duration = segment_duration / len(chunks)
+        for chunk_index, chunk in enumerate(chunks):
+            chunk_start = start + chunk_duration * chunk_index
+            chunk_end = end if chunk_index == len(chunks) - 1 else start + chunk_duration * (chunk_index + 1)
+            blocks.append(
+                "\n".join(
+                    [
+                        str(subtitle_index),
+                        f"{format_srt_time(chunk_start)} --> {format_srt_time(chunk_end)}",
+                        chunk,
+                    ]
+                )
+            )
+            subtitle_index += 1
+
+    return blocks
+
+
+def create_script_subtitle_file(
+    task_id: str,
+    list_script: Sequence[dict],
+    output_file: Optional[str] = None,
+    include_ost: Optional[Iterable[int]] = None,
+    max_chars: int = DEFAULT_MAX_CHARS_PER_SUBTITLE,
+) -> str:
+    """Create a full SRT file from script narration and edited timeline ranges."""
+    if not list_script:
+        return ""
+
+    if include_ost is None:
+        include_ost = DEFAULT_SUBTITLE_OST_TYPES
+
+    blocks = _build_srt_blocks(list_script, include_ost=include_ost, max_chars=max_chars)
+    if not blocks:
+        logger.warning("程序化字幕未生成内容")
+        return ""
+
+    if output_file is None:
+        output_file = os.path.join(utils.task_dir(task_id), "script_subtitles.srt")
+
+    output_dir = os.path.dirname(output_file)
+    if output_dir:
+        os.makedirs(output_dir, exist_ok=True)
+    with open(output_file, "w", encoding="utf-8") as f:
+        f.write("\n\n".join(blocks))
+        f.write("\n")
+
+    logger.info(f"程序化字幕生成成功: {output_file}, 共 {len(blocks)} 条")
+    return output_file

app/services/task.py

@@ -10,7 +10,16 @@ from app.config import config
 from app.config.audio_config import AudioConfig, get_recommended_volumes_for_content
 from app.models import const
 from app.models.schema import VideoClipParams
-from app.services import (voice, audio_merger, subtitle_merger, clip_video, merger_video, update_script, generate_video)
+from app.services import (
+    voice,
+    audio_merger,
+    subtitle_merger,
+    clip_video,
+    merger_video,
+    update_script,
+    generate_video,
+    script_subtitle,
+)
 from app.services import state as sm
 from app.utils import utils
 
@@ -561,8 +570,20 @@ def start_subclip_unified(task_id: str, params: VideoClipParams):
             )
             logger.info(f"音频文件合并成功->{merged_audio_path}")
 
-            # 合并字幕文件
+            # 优先基于脚本文案和成片时间线生成字幕，失败时回退到TTS字幕合并
-            merged_subtitle_path = subtitle_merger.merge_subtitle_files(new_script_list)
+            merged_subtitle_path = ""
+            if getattr(params, "subtitle_enabled", True):
+                try:
+                    merged_subtitle_path = script_subtitle.create_script_subtitle_file(
+                        task_id=task_id,
+                        list_script=new_script_list,
+                    )
+                except Exception as e:
+                    logger.warning(f"程序化字幕生成失败，将尝试合并TTS字幕: {e}")
+
+            if not merged_subtitle_path and getattr(params, "subtitle_enabled", True):
+                merged_subtitle_path = subtitle_merger.merge_subtitle_files(new_script_list)
+
             if merged_subtitle_path:
                 logger.info(f"字幕文件合并成功->{merged_subtitle_path}")
             else:
@@ -630,7 +651,9 @@ def start_subclip_unified(task_id: str, params: VideoClipParams):
     6. 合并字幕/BGM/配音/视频
     """
     output_video_path = path.join(utils.task_dir(task_id), f"combined.mp4")
-    auto_transcription_enabled = _is_auto_transcription_enabled(params)
+    auto_transcription_enabled = _is_auto_transcription_enabled(params) and not bool(merged_subtitle_path)
+    if _is_auto_transcription_enabled(params) and merged_subtitle_path:
+        logger.info("已生成字幕文件，跳过最终视频自动转录")
     merge_output_video_path = (
         path.join(utils.task_dir(task_id), "combined_without_auto_subtitles.mp4")
         if auto_transcription_enabled

app/services/test_script_subtitle_unittest.py

@@ -0,0 +1,94 @@
+import tempfile
+import unittest
+from pathlib import Path
+
+from app.services import script_subtitle
+
+
+class ScriptSubtitleTests(unittest.TestCase):
+    def test_split_narration_prefers_punctuation_boundaries(self):
+        chunks = script_subtitle.split_narration(
+            "她终于意识到，这场婚姻不是爱情，而是一场交易。",
+            max_chars=12,
+        )
+
+        self.assertEqual(
+            ["她终于意识到", "这场婚姻不是爱情", "而是一场交易"],
+            chunks,
+        )
+
+    def test_time_range_parsing_supports_milliseconds(self):
+        start, end = script_subtitle.parse_time_range("00:00:01,500-00:00:03,250")
+
+        self.assertAlmostEqual(1.5, start)
+        self.assertAlmostEqual(3.25, end)
+
+    def test_create_script_subtitle_file_skips_original_audio_segments(self):
+        list_script = [
+            {
+                "_id": 1,
+                "OST": 0,
+                "narration": "第一句解说。第二句解说。",
+                "editedTimeRange": "00:00:00-00:00:04",
+                "duration": 4,
+            },
+            {
+                "_id": 2,
+                "OST": 1,
+                "narration": "这句是原声，不应该默认生成。",
+                "editedTimeRange": "00:00:04-00:00:08",
+                "duration": 4,
+            },
+            {
+                "_id": 3,
+                "OST": 2,
+                "narration": "混合片段也保留解说字幕。",
+                "editedTimeRange": "00:00:08-00:00:12",
+                "duration": 4,
+            },
+        ]
+
+        with tempfile.TemporaryDirectory() as temp_dir:
+            output_file = Path(temp_dir) / "script_subtitles.srt"
+            result = script_subtitle.create_script_subtitle_file(
+                task_id="test",
+                list_script=list_script,
+                output_file=str(output_file),
+                max_chars=16,
+            )
+
+            self.assertEqual(str(output_file), result)
+            content = output_file.read_text(encoding="utf-8")
+
+        self.assertIn("00:00:00,000 -->", content)
+        self.assertIn("第一句解说", content)
+        self.assertIn("混合片段也保留解说字幕", content)
+        self.assertNotIn("这句是原声", content)
+        self.assertNotIn("。", content)
+        self.assertNotIn("，", content)
+
+    def test_create_script_subtitle_file_uses_duration_when_edited_range_missing(self):
+        list_script = [
+            {
+                "_id": 1,
+                "OST": 0,
+                "narration": "没有 editedTimeRange 时使用 duration。",
+                "duration": 3,
+            }
+        ]
+
+        with tempfile.TemporaryDirectory() as temp_dir:
+            output_file = Path(temp_dir) / "script_subtitles.srt"
+            script_subtitle.create_script_subtitle_file(
+                task_id="test",
+                list_script=list_script,
+                output_file=str(output_file),
+            )
+            content = output_file.read_text(encoding="utf-8")
+
+        self.assertIn("00:00:00,000 -->", content)
+        self.assertIn("--> 00:00:03,000", content)
+
+
+if __name__ == "__main__":
+    unittest.main()

webui.py

@@ -181,6 +181,11 @@ def render_generate_button():
                     div[data-testid="stDialog"] div[data-testid="stProgress"] {
                         margin-bottom: 0.75rem;
                     }
+                    div[data-testid="stDialog"] video {
+                        max-height: 62vh;
+                        object-fit: contain;
+                        background: #000;
+                    }
                 </style>
                 """,
                 unsafe_allow_html=True,
@@ -246,8 +251,16 @@ def render_generate_button():
                         video_files = task.get("videos", [])
                         try:
                             if video_files:
+                                aspect = getattr(params, "video_aspect", "")
+                                aspect = getattr(aspect, "value", aspect)
+                                preview_width = 320 if aspect in {
+                                    VideoAspect.portrait.value,
+                                    VideoAspect.portrait_2.value,
+                                } else 600
                                 for url in video_files:
-                                    st.video(url)
+                                    _, preview_col, _ = st.columns([1, 2, 1])
+                                    with preview_col:
+                                        st.video(url, width=preview_width)
                         except Exception as e:
                             logger.error(f"播放视频失败: {e}")