fix: 标题顶部边距60→120，标题位置下移 (#1506)

Co-authored-by: xiaoxia <dev@xiaoxiajianji.com>
Co-committed-by: xiaoxia <dev@xiaoxiajianji.com>

apps/api/app/api/routes/tts.py

@@ -381,15 +381,34 @@ def preview_tts(
     request: TTSPreviewRequest,
     authenticated_user: AuthenticatedUser = Depends(get_current_user),
     cosyvoice_service: CosyVoiceService = Depends(get_cosyvoice_service),
+    voice_clone_repo=Depends(get_voice_clone_profile_repository),
 ) -> TTSPreviewResponse:
     """TTS 预览（试听）——同步合成，立即返回音频 URL。
 
     用于前端预览配音效果，限制文本长度 200 字以内。
+    支持预设音色和克隆音色：克隆音色传的是 profile UUID，需解析为 CosyVoice voice_id。
     """
+    # 解析 voice_id：前端可能传 VoiceCloneProfile UUID 或预设音色 ID
+    actual_voice_id = request.voice_id
+    profile = voice_clone_repo.get(request.voice_id)
+    if profile is not None:
+        # 命中克隆音色 profile — 校验归属权限
+        if profile.user_id != authenticated_user.user.id:
+            raise HTTPException(
+                status_code=status.HTTP_403_FORBIDDEN,
+                detail="无权访问该音色",
+            )
+        if not profile.voice_id:
+            raise HTTPException(
+                status_code=status.HTTP_400_BAD_REQUEST,
+                detail="音色克隆尚未完成，请稍后再试",
+            )
+        actual_voice_id = profile.voice_id
+
     try:
         result = cosyvoice_service.synthesize_speech(
             text=request.text,
-            voice_id=request.voice_id,
+            voice_id=actual_voice_id,
             speed=request.speed,
         )
     except CosyVoiceError as e:

apps/web/src/pages/generate/components/FrontendPreviewPlayer.tsx

@@ -76,12 +76,59 @@ function buildPlaybackSegments(
 const FrontendPreviewPlayer: React.FC<FrontendPreviewPlayerProps> = ({
   assets,
   template,
-  videoRatio: _videoRatio,
+  videoRatio,
   ready,
   voiceAudioUrl,
   titleSettings,
 }) => {
   const segments = useMemo(() => buildPlaybackSegments(assets, template), [assets, template])
+
+  // ── ASS 坐标系参数（与后端 ass_subtitle_builder.py 一致） ──
+  const TITLE_MARGIN_TOP = 120
+  const TITLE_MARGIN_BOTTOM = 60
+  const TITLE_MARGIN_SIDE = 40
+  const playRes = (() => {
+    switch (videoRatio) {
+      case "16:9":
+        return { width: 1920, height: 1080 }
+      case "1:1":
+        return { width: 1080, height: 1080 }
+      case "9:16":
+      default:
+        return { width: 1080, height: 1920 }
+    }
+  })()
+  const playerContainerRef = useRef<HTMLDivElement>(null)
+  const [containerHeight, setContainerHeight] = useState(0)
+  useEffect(() => {
+    const el = playerContainerRef.current
+    if (!el) return
+    const ro = new ResizeObserver((entries) => {
+      for (const entry of entries) {
+        const h = entry.contentRect.height
+        if (h > 0) setContainerHeight(h)
+      }
+    })
+    ro.observe(el)
+    const rect = el.getBoundingClientRect()
+    if (rect.height > 0) setContainerHeight(rect.height)
+    return () => ro.disconnect()
+  }, [])
+
+  // 标题字号按容器高度与 PlayResY 的比例缩放
+  const titleFontSizePx =
+    containerHeight > 0
+      ? ((titleSettings?.size ?? 36) / playRes.height) * containerHeight
+      : (titleSettings?.size ?? 36)
+  const titleSidePct = (TITLE_MARGIN_SIDE / playRes.width) * 100
+  const titleTopPct = (TITLE_MARGIN_TOP / playRes.height) * 100
+  const titleBottomPct = (TITLE_MARGIN_BOTTOM / playRes.height) * 100
+  // 描边/阴影也要按缩放比例放大
+  const titleScale = containerHeight > 0 ? containerHeight / playRes.height : 1
+  const titleStrokeWidth = Math.max(1, 2 * titleScale)
+  const titleShadowBlur = 4 * titleScale
+  const titleShadowOffset = 2 * titleScale
+
   // 默认走原生 video 播放（浏览器硬件解码，独立线程，不阻塞 UI）
   // WebCodecs 仅在明确需要时启用（保留代码作为兜底）
   const useWebCodecs = false
@@ -374,6 +421,7 @@ const FrontendPreviewPlayer: React.FC<FrontendPreviewPlayerProps> = ({
 
   return (
     <div
+      ref={playerContainerRef}
       style={{
         position: "relative",
         width: "100%",
@@ -434,48 +482,55 @@ const FrontendPreviewPlayer: React.FC<FrontendPreviewPlayerProps> = ({
           />
         ))}
 
-      {/* 标题CSS叠加层 — video fallback 路径也要渲染 */}
+      {/* 标题CSS叠加层 — 与后端 ASS 烧录坐标系 1:1 对齐 */}
       {titleSettings?.title && (
         <div
           style={{
             position: "absolute",
-            left: 0,
-            right: 0,
+            inset: 0,
             zIndex: 5,
-            display: "flex",
-            justifyContent: "center",
             pointerEvents: "none",
-            ...(titleSettings.position === "top"
-              ? { top: "10%" }
-              : titleSettings.position === "center"
-                ? { top: "50%", transform: "translateY(-50%)" }
-                : { bottom: "15%" }),
+            overflow: "hidden",
           }}
         >
-          <span
+          <div
             style={{
-              fontSize: titleSettings.size,
-              fontFamily: titleSettings.font || "思源黑体",
-              color: titleSettings.color || "#ffffff",
-              fontWeight: titleSettings.bold ? 700 : 400,
-              fontStyle: titleSettings.italic ? "italic" : "normal",
-              textShadow: [
-                titleSettings.shadow ? "0 2px 8px rgba(0,0,0,0.7)" : undefined,
-                titleSettings.stroke
-                  ? "1px 1px 0 rgba(0,0,0,0.5), -1px -1px 0 rgba(0,0,0,0.5), 1px -1px 0 rgba(0,0,0,0.5), -1px 1px 0 rgba(0,0,0,0.5)"
-                  : undefined,
-                "0 1px 3px rgba(0,0,0,0.4)",
-              ]
-                .filter(Boolean)
-                .join(", "),
-              maxWidth: "90%",
+              position: "absolute",
+              left: `${titleSidePct}%`,
+              right: `${titleSidePct}%`,
               textAlign: "center",
-              lineHeight: 1.3,
-              wordBreak: "break-word",
+              ...(titleSettings.position === "top"
+                ? { top: `${titleTopPct}%` }
+                : titleSettings.position === "center"
+                  ? { top: "50%", transform: "translateY(-50%)" }
+                  : { bottom: `${titleBottomPct}%` }),
             }}
           >
-            {titleSettings.title}
-          </span>
+            <span
+              style={{
+                fontSize: `${titleFontSizePx}px`,
+                fontFamily: titleSettings.font || "思源黑体",
+                color: titleSettings.color || "#ffffff",
+                fontWeight: titleSettings.bold ? 700 : 400,
+                fontStyle: titleSettings.italic ? "italic" : "normal",
+                lineHeight: 1.3,
+                wordBreak: "break-word",
+                WebkitTextStroke: titleSettings.stroke
+                  ? `${titleStrokeWidth}px #000000`
+                  : undefined,
+                textShadow: titleSettings.shadow
+                  ? `${titleShadowOffset}px ${titleShadowOffset}px ${titleShadowBlur}px rgba(0,0,0,0.8)`
+                  : undefined,
+              }}
+            >
+              {titleSettings.title.split(/[/／]/).map((part, i) => (
+                <span key={i}>
+                  {i > 0 && <br />}
+                  {part}
+                </span>
+              ))}
+            </span>
+          </div>
         </div>
       )}
 

apps/web/src/pages/generate/components/PreviewVideoPanel.tsx

@@ -40,17 +40,33 @@ interface PreviewVideoPanelProps {
 }
 
 /* ── ASS 坐标系参数（与后端 ass_subtitle_builder.py 一致） ── */
-const ASS_VIDEO_HEIGHT = 720
-const ASS_TITLE_MARGIN_TOP = 60
-const ASS_TITLE_MARGIN_BOTTOM = 60
-const ASS_TITLE_MARGIN_SIDE = 40
+const TITLE_MARGIN_TOP = 120
+const TITLE_MARGIN_BOTTOM = 60
+const TITLE_MARGIN_SIDE = 40
 
-function getPositionStyle(position: string): React.CSSProperties {
-  const sidePercent = (ASS_TITLE_MARGIN_SIDE / 1280) * 100
+/** 根据视频比例返回后端实际渲染分辨率（PlayResX × PlayResY） */
+function getResolution(ratio: string): { width: number; height: number } {
+  switch (ratio) {
+    case "16:9":
+      return { width: 1920, height: 1080 }
+    case "1:1":
+      return { width: 1080, height: 1080 }
+    case "9:16":
+    default:
+      return { width: 1080, height: 1920 }
+  }
+}
+
+function getPositionStyle(
+  position: string,
+  playResX: number,
+  playResY: number,
+): React.CSSProperties {
+  const sidePercent = (TITLE_MARGIN_SIDE / playResX) * 100
   switch (position) {
     case "bottom":
       return {
-        bottom: `${(ASS_TITLE_MARGIN_BOTTOM / ASS_VIDEO_HEIGHT) * 100}%`,
+        bottom: `${(TITLE_MARGIN_BOTTOM / playResY) * 100}%`,
         left: `${sidePercent}%`,
         right: `${sidePercent}%`,
         textAlign: "center",
@@ -66,7 +82,7 @@ function getPositionStyle(position: string): React.CSSProperties {
     case "top":
     default:
       return {
-        top: `${(ASS_TITLE_MARGIN_TOP / ASS_VIDEO_HEIGHT) * 100}%`,
+        top: `${(TITLE_MARGIN_TOP / playResY) * 100}%`,
         left: `${sidePercent}%`,
         right: `${sidePercent}%`,
         textAlign: "center",
@@ -74,11 +90,16 @@ function getPositionStyle(position: string): React.CSSProperties {
   }
 }
 
-function buildTitleStyle(settings: TitleSettings, containerHeight: number): React.CSSProperties {
+function buildTitleStyle(
+  settings: TitleSettings,
+  containerHeight: number,
+  playResY: number,
+): React.CSSProperties {
+  // 字号按容器高度与 PlayResY 的比例缩放，不设上限（与后端一致）
   const fontSizePx =
     containerHeight > 0
-      ? (Math.min(settings.size, 96) / ASS_VIDEO_HEIGHT) * containerHeight
-      : (Math.min(settings.size, 96) / ASS_VIDEO_HEIGHT) * 400
+      ? (settings.size / playResY) * containerHeight
+      : (settings.size / playResY) * 400
 
   const base: React.CSSProperties = {
     fontFamily: getFontFamily(settings.font),
@@ -90,8 +111,6 @@ function buildTitleStyle(settings: TitleSettings, containerHeight: number): Reac
     wordBreak: "break-word",
     pointerEvents: "none",
     userSelect: "none",
-    paddingLeft: `${(ASS_TITLE_MARGIN_SIDE / 1280) * 100}%`,
-    paddingRight: `${(ASS_TITLE_MARGIN_SIDE / 1280) * 100}%`,
   }
   if (settings.stroke) base.WebkitTextStroke = "1px #000000"
   if (settings.shadow) base.textShadow = "2px 2px 4px rgba(0,0,0,0.8)"
@@ -99,7 +118,10 @@ function buildTitleStyle(settings: TitleSettings, containerHeight: number): Reac
 }
 
 /** CSS 标题实时预览覆盖层 */
-const TitleOverlay: React.FC<{ titleSettings: TitleSettings }> = ({ titleSettings }) => {
+const TitleOverlay: React.FC<{ titleSettings: TitleSettings; videoRatio: string }> = ({
+  titleSettings,
+  videoRatio,
+}) => {
   const containerRef = useRef<HTMLDivElement>(null)
   const [containerHeight, setContainerHeight] = useState(400)
 
@@ -118,12 +140,15 @@ const TitleOverlay: React.FC<{ titleSettings: TitleSettings }> = ({ titleSetting
     return () => ro.disconnect()
   }, [])
 
+  const { width: playResX, height: playResY } = getResolution(videoRatio)
+
   const positionStyle = useMemo(
-    () => getPositionStyle(titleSettings.position),
-    [titleSettings.position],
+    () => getPositionStyle(titleSettings.position, playResX, playResY),
+    // eslint-disable-next-line react-hooks/exhaustive-deps
+    [titleSettings.position, playResX, playResY],
   )
   const titleStyle = useMemo(
-    () => buildTitleStyle(titleSettings, containerHeight),
+    () => buildTitleStyle(titleSettings, containerHeight, playResY),
     // eslint-disable-next-line react-hooks/exhaustive-deps
     [
       containerHeight,
@@ -134,6 +159,7 @@ const TitleOverlay: React.FC<{ titleSettings: TitleSettings }> = ({ titleSetting
       titleSettings.italic,
       titleSettings.stroke,
       titleSettings.shadow,
+      playResY,
     ],
   )
 
@@ -151,7 +177,7 @@ const TitleOverlay: React.FC<{ titleSettings: TitleSettings }> = ({ titleSetting
       }}
     >
       <div style={{ ...positionStyle, ...titleStyle, position: "absolute" }}>
-        {displayTitle.split("/").map((part, i) => (
+        {displayTitle.split(/[/／]/).map((part, i) => (
           <span key={i}>
             {i > 0 && <br />}
             {part}
@@ -266,7 +292,9 @@ export const PreviewVideoPanel: React.FC<PreviewVideoPanelProps> = ({
         )}
 
         {/* 标题样式实时预览层（仅在有视频时叠加） */}
-        {isReady && titleSettings && <TitleOverlay titleSettings={titleSettings} />}
+        {isReady && titleSettings && (
+          <TitleOverlay titleSettings={titleSettings} videoRatio={videoRatio} />
+        )}
 
         {/* stale 遮罩：配置变更提示 */}
         {isStale && (

apps/web/src/pages/generate/components/title/TitleStylePanel.tsx

@@ -94,7 +94,7 @@ const TitleStylePanel: React.FC<TitleStylePanelProps> = ({
           className="xx-slider"
           type="range"
           min={12}
-          max={48}
+          max={128}
           value={settings.size}
           onChange={(e) => onUpdateSize(Number(e.target.value))}
         />

apps/worker/video_processing/render_audio.py

@@ -512,6 +512,9 @@ def mix_with_independent_audio(
                 clip_filters.append("asetpts=PTS-STARTPTS")
             else:
                 clip_filters.append("asetpts=PTS-STARTPTS")
+            vol = _clip_volume(clip)
+            if abs(vol - 1.0) >= 1e-6:
+                clip_filters.append(f"volume={vol:.4f}")
             # aformat 归一化：concat/amix 前统一音频参数，否则不同采样率/声道会失败
             clip_filters.append(AFORMAT)
             filter_parts.append(f"[{input_idx}:a]{','.join(clip_filters)}[ma{input_idx}]")

apps/worker/video_processing/unified_render_service.py

@@ -824,6 +824,17 @@ class UnifiedRenderService:
                 )
                 audio_layer.clips.append(vo_clip)
 
+            # replace 模式：静音原视频音轨（main + broll 图层）
+            if tts_config.overlap_mode == "replace":
+                for layer in layers:
+                    if layer.role in ("main", "broll"):
+                        for clip in layer.clips:
+                            clip.config["volume"] = 0
+                logger.info(
+                    "TTS replace 模式：已静音原视频音轨: plan_id=%s",
+                    self.plan.id,
+                )
+
             logger.info(
                 "TTS 配音已添加: plan_id=%s voice_id=%s segments=%d total_%.2fs",
                 self.plan.id,
@@ -889,6 +900,16 @@ class UnifiedRenderService:
             )
             audio_layer.clips.append(vo_clip)
 
+            # 配音素材库默认替换原音：静音原视频音轨（main + broll 图层）
+            for layer in layers:
+                if layer.role in ("main", "broll"):
+                    for clip in layer.clips:
+                        clip.config["volume"] = 0
+            logger.info(
+                "配音素材库：已静音原视频音轨: plan_id=%s",
+                self.plan.id,
+            )
+
             logger.info(
                 "配音素材库音频已添加到 audio 图层: plan_id=%s duration=%.2fs",
                 self.plan.id,

packages/domain/ass_subtitle_builder.py

@@ -20,7 +20,7 @@ logger = logging.getLogger(__name__)
 # ── 常量 ──────────────────────────────────────────────────────────────────────
 
 # Title/Subtitle 默认边距（像素）
-TITLE_MARGIN_TOP = 60
+TITLE_MARGIN_TOP = 120
 TITLE_MARGIN_BOTTOM = 60
 TITLE_MARGIN_SIDE = 40
 
@@ -147,6 +147,8 @@ def escape_ass_text(text: str) -> str:
     Returns:
         转义后的 ASS 文本
     """
+    # 用户手动换行符（半角/全角斜杠）转为 ASS 硬换行（在自动换行之前优先处理）
+    text = text.replace("／", "\\N").replace("/", "\\N")
     # 将实际换行转为 ASS 硬换行
     text = text.replace("\r\n", "\\N").replace("\n", "\\N").replace("\r", "\\N")
     # 转义大括号（ASS 用它做样式覆盖标签）
@@ -195,26 +197,33 @@ def _wrap_title_text(
     if available_width <= 0:
         return text
 
-    lines: list[str] = []
-    current_line = ""
-    current_width = 0.0
+    # 先按已有 \N 分段，每段独立自动换行，最后用 \N 拼回
+    segments = text.split("\\N")
+    wrapped_segments: list[str] = []
 
-    for ch in text:
-        # CJK 字符按全角估算，其他按半角
-        char_width = float(font_size) if ord(ch) > 0x2E80 else font_size * 0.55
+    for seg in segments:
+        lines: list[str] = []
+        current_line = ""
+        current_width = 0.0
 
-        if current_width + char_width > available_width and current_line:
+        for ch in seg:
+            # CJK 字符按全角估算，其他按半角
+            char_width = float(font_size) if ord(ch) > 0x2E80 else font_size * 0.55
+
+            if current_width + char_width > available_width and current_line:
+                lines.append(current_line)
+                current_line = ch
+                current_width = char_width
+            else:
+                current_line += ch
+                current_width += char_width
+
+        if current_line:
             lines.append(current_line)
-            current_line = ch
-            current_width = char_width
-        else:
-            current_line += ch
-            current_width += char_width
 
-    if current_line:
-        lines.append(current_line)
+        wrapped_segments.append("\\N".join(lines))
 
-    return "\\N".join(lines)
+    return "\\N".join(wrapped_segments)
 
 
 def build_ass_content(
@@ -247,6 +256,12 @@ def build_ass_content(
     title_config = title_config or {}
     subtitle_config = subtitle_config or {}
 
+    # ── 字段名归一化：前端传 font_size/font_color，内部用 size/color ──
+    if "font_size" in title_config and "size" not in title_config:
+        title_config["size"] = title_config["font_size"]
+    if "font_color" in title_config and "color" not in title_config:
+        title_config["color"] = title_config["font_color"]
+
     # ── 兼容前端简化格式：stroke/shadow 为 boolean 时，转换为标准 dict ──
     # 前端 TitleSettings 发送 stroke=true/false, shadow=true/false
     # 后端 build_ass_style 期望 stroke={enabled, color, width}, shadow={enabled, blur, offset_x, offset_y}
@@ -296,7 +311,7 @@ def build_ass_content(
             build_ass_style(
                 "TitleStyle",
                 font_name=title_config.get("font", "思源黑体"),
-                font_size=min(int(title_config.get("size", 36)), 36),
+                font_size=int(title_config.get("size", 36)),
                 primary_color=title_color,
                 outline_color=stroke_color,
                 outline_width=stroke_width,
@@ -313,7 +328,7 @@ def build_ass_content(
 
         # 根据视频宽度和字号自动换行标题，防止超出画面
         # 先 escape 特殊字符，再插入换行符 \N，避免顺序颠倒导致 \N 被转义
-        title_font_size = min(int(title_config.get("size", 36)), 36)
+        title_font_size = int(title_config.get("size", 36))
         safe_title_text_raw = escape_ass_text(title_text)
         safe_title_text = _wrap_title_text(safe_title_text_raw, video_width, title_font_size)
 

tests/unit/domain/test_ass_subtitle_builder.py

@@ -218,6 +218,14 @@ class TestEscapeAssText:
     def test_chinese_text(self):
         assert escape_ass_text("你好世界") == "你好世界"
 
+    def test_slash_converted_to_newline(self):
+        """半角斜杠 / 应转为 ASS 硬换行。"""
+        assert escape_ass_text("标题一/标题二") == "标题一\\N标题二"
+
+    def test_fullwidth_slash_converted_to_newline(self):
+        """全角斜杠 ／ 应转为 ASS 硬换行。"""
+        assert escape_ass_text("标题一／标题二") == "标题一\\N标题二"
+
     def test_backslash_n_in_input(self):
         # 文本里本身有 \n 字符串（不是换行符）
         result = escape_ass_text("\\n")
@@ -408,11 +416,53 @@ class TestBuildAssContent:
             title_text="T",
             title_config={"size": 72},
         )
-        # 在TitleStyle行里查找字体大小
+        # 在TitleStyle行里查找字体大小（字号上限已移除，72应原样保留）
         for line in result.split("\n"):
             if line.startswith("Style: TitleStyle"):
                 parts = line.split(",")
-                assert parts[2] == "36"
+                assert parts[2] == "72"
+                break
+
+    def test_title_font_size_frontend_field_alias(self):
+        """前端传 font_size 应归一化为内部 size 字段。"""
+        result = build_ass_content(
+            video_width=1920,
+            video_height=1080,
+            video_duration=5.0,
+            title_text="T",
+            title_config={"font_size": 48},
+        )
+        for line in result.split("\n"):
+            if line.startswith("Style: TitleStyle"):
+                parts = line.split(",")
+                assert parts[2] == "48"
+                break
+
+    def test_title_font_color_frontend_field_alias(self):
+        """前端传 font_color 应归一化为内部 color 字段。"""
+        result = build_ass_content(
+            video_width=1920,
+            video_height=1080,
+            video_duration=5.0,
+            title_text="T",
+            title_config={"font_color": "#FF0000"},
+        )
+        # 红色 → &H0000FF
+        assert "&H0000FF" in result
+
+    def test_title_size_takes_precedence_over_font_size(self):
+        """同时传 size 和 font_size 时，size 优先。"""
+        result = build_ass_content(
+            video_width=1920,
+            video_height=1080,
+            video_duration=5.0,
+            title_text="T",
+            title_config={"size": 56, "font_size": 28},
+        )
+        for line in result.split("\n"):
+            if line.startswith("Style: TitleStyle"):
+                parts = line.split(",")
+                assert parts[2] == "56"
                 break
 
     def test_title_bold(self):

tests/unit/test_ass_subtitle_builder.py

@@ -190,6 +190,19 @@ class TestEscapeAssText:
     def test_chinese_text(self):
         assert escape_ass_text("你好世界") == "你好世界"
 
+    def test_slash_converted_to_newline(self):
+        """半角斜杠 / 应转为 ASS 硬换行。"""
+        assert escape_ass_text("第一行/第二行") == "第一行\\N第二行"
+
+    def test_fullwidth_slash_converted_to_newline(self):
+        """全角斜杠 ／ 应转为 ASS 硬换行。"""
+        assert escape_ass_text("第一行／第二行") == "第一行\\N第二行"
+
+    def test_mixed_slashes_and_newlines(self):
+        """斜杠和换行符都应转为硬换行。"""
+        result = escape_ass_text("A/B\nC／D")
+        assert result == "A\\NB\\NC\\ND"
+
 
 # ── 时间格式化 ────────────────────────────────────────────────────────────────
 
@@ -440,7 +453,7 @@ class TestBuildAssContent:
 
 class TestConstants:
     def test_title_margin_top(self):
-        assert TITLE_MARGIN_TOP == 60
+        assert TITLE_MARGIN_TOP == 120
 
     def test_title_margin_bottom(self):
         assert TITLE_MARGIN_BOTTOM == 60
@@ -491,6 +504,35 @@ class TestWrapTitleText:
         """字号为0时直接返回原文。"""
         assert _wrap_title_text("测试", 1080, 0) == "测试"
 
+    def test_preserves_explicit_newline(self):
+        """已有的 \\N 换行标记应保留，不被当普通字符算宽度。"""
+        text = "第一行\\N第二行"
+        result = _wrap_title_text(text, video_width=1080, font_size=48)
+        assert result == text
+
+    def test_explicit_newline_each_segment_wraps_independently(self):
+        """\\N 分段后，每段各自自动换行。"""
+        # 480px 宽，48px 字号，可用 400px，每段约8个中文字
+        text = "这是第一段很长很长很长的内容\\N这是第二段也很长很长的内容"
+        result = _wrap_title_text(text, video_width=480, font_size=48)
+        # 应该有多个 \N：用户手动的 + 自动换行的
+        assert "\\N" in result
+        segments = result.split("\\N")
+        # 至少3行（两段都需要换行）
+        assert len(segments) >= 3
+        # 验证包含两段的文字
+        joined = result.replace("\\N", "")
+        assert "第一段" in joined
+        assert "第二段" in joined
+
+    def test_multiple_explicit_newlines(self):
+        """多个 \\N 分段都应保留。"""
+        text = "A\\NB\\NC"
+        result = _wrap_title_text(text, video_width=1080, font_size=48)
+        assert result == text
+        assert result.count("\\N") == 2
+
+
     def test_build_ass_content_integration(self):
         """集成测试：build_ass_content 中的标题应该自动换行。"""
         long_title = "这是一段非常长的标题文字用于测试自动换行功能是否正常工作"

tests/unit/test_render_audio_pure.py

@@ -11,6 +11,7 @@ from video_processing.render_audio import (
     RenderContext,
     clip_effective_duration,
     clip_has_audio,
+    mix_with_independent_audio,
 )
 from video_processing.unified_render_service import ResolvedClip
 
@@ -153,3 +154,93 @@ class TestRenderContext:
         """音频缓存初始为空."""
         ctx = _make_ctx()
         assert ctx._audio_cache == {}
+
+
+class TestMixWithIndependentAudioVolume:
+    """测试 mix_with_independent_audio 中主视频 clip 音量滤镜是否生效。"""
+
+    def _make_clip(self, volume=None, clip_id="c1", duration=5.0):
+        """创建测试用 ResolvedClip。"""
+        config = {}
+        if volume is not None:
+            config["volume"] = volume
+        return ResolvedClip(
+            clip_id=clip_id,
+            asset_id=f"a_{clip_id}",
+            local_path=Path(f"/tmp/{clip_id}.mp4"),
+            clip_type="main",
+            order=0,
+            duration=duration,
+            actual_duration=duration,
+            config=config,
+        )
+
+    def test_main_clip_volume_zero_applied_in_filter(self):
+        """volume=0 的主视频 clip 应在 filter_complex 中包含 volume=0.0000。"""
+        from unittest.mock import patch
+
+        ctx = _make_ctx()
+        main_clip = self._make_clip(volume=0, clip_id="m1")
+        captured_cmd = {}
+
+        def fake_run_ffmpeg(cmd):
+            captured_cmd["cmd"] = cmd
+
+        with patch("video_processing.render_audio.run_ffmpeg", side_effect=fake_run_ffmpeg):
+            mix_with_independent_audio(
+                ctx=ctx,
+                main_clips=[main_clip],
+                audio_clips=[],
+                output_path=Path("/tmp/out.m4a"),
+                video_duration=5.0,
+            )
+
+        filter_complex = captured_cmd["cmd"][captured_cmd["cmd"].index("-filter_complex") + 1]
+        assert "volume=0.0000" in filter_complex, f"volume filter missing in: {filter_complex}"
+
+    def test_main_clip_default_volume_no_filter(self):
+        """默认 volume=1.0 时不应添加 volume 滤镜。"""
+        from unittest.mock import patch
+
+        ctx = _make_ctx()
+        main_clip = self._make_clip(clip_id="m1")  # no volume set
+        captured_cmd = {}
+
+        def fake_run_ffmpeg(cmd):
+            captured_cmd["cmd"] = cmd
+
+        with patch("video_processing.render_audio.run_ffmpeg", side_effect=fake_run_ffmpeg):
+            mix_with_independent_audio(
+                ctx=ctx,
+                main_clips=[main_clip],
+                audio_clips=[],
+                output_path=Path("/tmp/out.m4a"),
+                video_duration=5.0,
+            )
+
+        filter_complex = captured_cmd["cmd"][captured_cmd["cmd"].index("-filter_complex") + 1]
+        # 默认音量不应出现 volume= 滤镜
+        assert "volume=" not in filter_complex, f"unexpected volume filter in: {filter_complex}"
+
+    def test_main_clip_partial_volume_applied(self):
+        """volume=0.5 的主视频 clip 应包含 volume=0.5000。"""
+        from unittest.mock import patch
+
+        ctx = _make_ctx()
+        main_clip = self._make_clip(volume=0.5, clip_id="m1")
+        captured_cmd = {}
+
+        def fake_run_ffmpeg(cmd):
+            captured_cmd["cmd"] = cmd
+
+        with patch("video_processing.render_audio.run_ffmpeg", side_effect=fake_run_ffmpeg):
+            mix_with_independent_audio(
+                ctx=ctx,
+                main_clips=[main_clip],
+                audio_clips=[],
+                output_path=Path("/tmp/out.m4a"),
+                video_duration=5.0,
+            )
+
+        filter_complex = captured_cmd["cmd"][captured_cmd["cmd"].index("-filter_complex") + 1]
+        assert "volume=0.5000" in filter_complex, f"volume filter missing in: {filter_complex}"

tests/unit/test_title_render_consistency.py

@@ -1,7 +1,7 @@
 """标题渲染前后端一致性测试。
 
 验证 build_ass_content 生成的 ASS 样式参数与前端 drawTitleOnCanvas.ts 一致：
-- 字号上限 36px
+- 字号不再设置上限，由前端/调用方控制
 - 描边宽度 2px
 - 阴影 blur=4, offset=2
 - boolean stroke/shadow 自动转换
@@ -10,14 +10,16 @@
 import sys
 from pathlib import Path
 
+import pytest
+
 # Ensure packages is importable
 sys.path.insert(0, str(Path(__file__).resolve().parents[2] / "packages"))
 
 from domain.ass_subtitle_builder import build_ass_content, build_ass_style
 
 
-class TestFontSizeCap:
-    """字号上限应与前端 Math.min(settings.size, 36) 一致。"""
+class TestFontSize:
+    """字号处理：默认值与保留逻辑，不再做上限截断。"""
 
     def test_default_font_size_is_36(self):
         """无 size 字段时，默认字号应为 36。"""
@@ -43,8 +45,8 @@ class TestFontSizeCap:
         )
         assert ",32," in content
 
-    def test_size_60_capped_at_36(self):
-        """size=60 应被 cap 到 36。"""
+    def test_size_60_preserved(self):
+        """size=60 应原样保留（字号上限已移除）。"""
         config = {"size": 60}
         content = build_ass_content(
             video_width=1080,
@@ -53,14 +55,40 @@ class TestFontSizeCap:
             title_text="测试标题",
             title_config=config,
         )
-        # 解析 Style 行的 Fontsize 字段（第3个字段，索引2）
         style_line = [line for line in content.splitlines() if line.startswith("Style: TitleStyle")][0]
         fields = [f.strip() for f in style_line.split(",")]
         font_size = int(fields[2])
-        assert font_size == 36, f"字号60应被cap到36, 实际={font_size}"
+        assert font_size == 60, f"字号60应原样保留, 实际={font_size}"
+
+    def test_font_size_alias_normalized(self):
+        """前端传 font_size 应归一化为 size。"""
+        config = {"font_size": 52}
+        content = build_ass_content(
+            video_width=1080,
+            video_height=1920,
+            video_duration=10.0,
+            title_text="测试标题",
+            title_config=config,
+        )
+        style_line = [line for line in content.splitlines() if line.startswith("Style: TitleStyle")][0]
+        fields = [f.strip() for f in style_line.split(",")]
+        assert fields[2] == "52", f"font_size=52 应归一化, 实际={fields[2]}"
+
+    def test_font_color_alias_normalized(self):
+        """前端传 font_color 应归一化为 color。"""
+        config = {"font_color": "#00FF00"}
+        content = build_ass_content(
+            video_width=1080,
+            video_height=1920,
+            video_duration=10.0,
+            title_text="测试标题",
+            title_config=config,
+        )
+        # 绿色 → &H00FF00
+        assert "&H00FF00" in content
 
     def test_size_24_preserved(self):
-        """size=24 应原样使用（小于36，不cap）。"""
+        """size=24 应原样使用。"""
         config = {"size": 24}
         content = build_ass_content(
             video_width=1080,
@@ -203,7 +231,7 @@ class TestFullStyleConsistency:
 
         # Fontname
         assert fields[1] == "思源黑体"
-        # Fontsize = 28 (小于36，不cap)
+        # Fontsize = 28
         assert fields[2] == "28"
         # Bold = -1 (True)
         assert fields[7] == "-1"
@@ -213,3 +241,39 @@ class TestFullStyleConsistency:
         assert int(fields[17]) == 2
         # Alignment = 8 (top)
         assert int(fields[18]) == 8
+
+
+class TestTitleSlashNewline:
+    """用户输入 / 或 ／ 应触发标题换行。"""
+
+    def test_halfwidth_slash_in_title(self):
+        content = build_ass_content(
+            video_width=1080,
+            video_height=1920,
+            video_duration=10.0,
+            title_text="第一行/第二行",
+            title_config={"size": 48},
+        )
+        for line in content.splitlines():
+            if "Dialogue" in line and "TitleStyle" in line:
+                assert "\\N" in line, f"斜杠应转为换行: {line}"
+                assert "第一行" in line
+                assert "第二行" in line
+                break
+        else:
+            pytest.fail("未找到 TitleStyle Dialogue 行")
+
+    def test_fullwidth_slash_in_title(self):
+        content = build_ass_content(
+            video_width=1080,
+            video_height=1920,
+            video_duration=10.0,
+            title_text="第一行／第二行",
+            title_config={"size": 48},
+        )
+        for line in content.splitlines():
+            if "Dialogue" in line and "TitleStyle" in line:
+                assert "\\N" in line, f"全角斜杠应转为换行: {line}"
+                break
+        else:
+            pytest.fail("未找到 TitleStyle Dialogue 行")

tests/unit/test_tts_preview.py

@@ -81,6 +81,12 @@ class TestTTSPreviewEndpoint:
         )
         app.dependency_overrides[get_cosyvoice_service] = lambda: mock_service
 
+        from app.dependencies import get_voice_clone_profile_repository
+
+        mock_clone_repo = MagicMock()
+        mock_clone_repo.get.return_value = None  # no profile found = preset voice
+        app.dependency_overrides[get_voice_clone_profile_repository] = lambda: mock_clone_repo
+
         client = self._make_client(app)
         resp = client.post(
             "/tts/preview",
@@ -120,6 +126,12 @@ class TestTTSPreviewEndpoint:
         )
         app.dependency_overrides[get_cosyvoice_service] = lambda: mock_service
 
+        from app.dependencies import get_voice_clone_profile_repository
+
+        mock_clone_repo = MagicMock()
+        mock_clone_repo.get.return_value = None  # no profile found = preset voice
+        app.dependency_overrides[get_voice_clone_profile_repository] = lambda: mock_clone_repo
+
         client = self._make_client(app)
         resp = client.post(
             "/tts/preview",
@@ -158,6 +170,12 @@ class TestTTSPreviewEndpoint:
         mock_service.synthesize_speech.side_effect = CosyVoiceError("API timeout")
         app.dependency_overrides[get_cosyvoice_service] = lambda: mock_service
 
+        from app.dependencies import get_voice_clone_profile_repository
+
+        mock_clone_repo = MagicMock()
+        mock_clone_repo.get.return_value = None  # no profile found = preset voice
+        app.dependency_overrides[get_voice_clone_profile_repository] = lambda: mock_clone_repo
+
         client = self._make_client(app)
         resp = client.post(
             "/tts/preview",
@@ -186,6 +204,12 @@ class TestTTSPreviewEndpoint:
         mock_service.synthesize_speech.side_effect = ValueError("text 不能为空")
         app.dependency_overrides[get_cosyvoice_service] = lambda: mock_service
 
+        from app.dependencies import get_voice_clone_profile_repository
+
+        mock_clone_repo = MagicMock()
+        mock_clone_repo.get.return_value = None  # no profile found = preset voice
+        app.dependency_overrides[get_voice_clone_profile_repository] = lambda: mock_clone_repo
+
         client = self._make_client(app)
         resp = client.post(
             "/tts/preview",
@@ -213,6 +237,12 @@ class TestTTSPreviewEndpoint:
         mock_service = MagicMock()
         app.dependency_overrides[get_cosyvoice_service] = lambda: mock_service
 
+        from app.dependencies import get_voice_clone_profile_repository
+
+        mock_clone_repo = MagicMock()
+        mock_clone_repo.get.return_value = None  # no profile found = preset voice
+        app.dependency_overrides[get_voice_clone_profile_repository] = lambda: mock_clone_repo
+
         client = self._make_client(app)
         resp = client.post(
             "/tts/preview",
@@ -239,9 +269,173 @@ class TestTTSPreviewEndpoint:
         mock_service = MagicMock()
         app.dependency_overrides[get_cosyvoice_service] = lambda: mock_service
 
+        from app.dependencies import get_voice_clone_profile_repository
+
+        mock_clone_repo = MagicMock()
+        mock_clone_repo.get.return_value = None  # no profile found = preset voice
+        app.dependency_overrides[get_voice_clone_profile_repository] = lambda: mock_clone_repo
+
         client = self._make_client(app)
         resp = client.post(
             "/tts/preview",
             json={"text": "hello", "voice_id": ""},
         )
         assert resp.status_code == 422
+
+    def test_preview_clone_voice_resolves_to_cosyvoice_id(self):
+        """Clone voice UUID is resolved to CosyVoice voice_id."""
+        from dataclasses import dataclass
+
+        from fastapi import FastAPI
+
+        app = FastAPI()
+        from app.api.routes.tts import router
+
+        app.include_router(router, prefix="/tts")
+
+        from app.auth import get_current_user
+        from app.dependencies import get_cosyvoice_service, get_voice_clone_profile_repository
+
+        fake_user = MagicMock()
+        fake_user.user.id = "user-1"
+        app.dependency_overrides[get_current_user] = lambda: fake_user
+
+        mock_service = MagicMock()
+        mock_service.synthesize_speech.return_value = FakeSynthesizeResult(
+            audio_url="https://x.com/cloned.mp3",
+            duration=1.8,
+        )
+        app.dependency_overrides[get_cosyvoice_service] = lambda: mock_service
+
+        # Mock voice clone profile with voice_id
+        mock_profile = MagicMock()
+        mock_profile.user_id = "user-1"
+        mock_profile.voice_id = "cosyvoice_actual_voice_123"
+        mock_clone_repo = MagicMock()
+        mock_clone_repo.get.return_value = mock_profile
+        app.dependency_overrides[get_voice_clone_profile_repository] = lambda: mock_clone_repo
+
+        client = self._make_client(app)
+        # Frontend sends the profile UUID as voice_id
+        resp = client.post(
+            "/tts/preview",
+            json={"text": "克隆音色测试", "voice_id": "abc123-uuid-of-profile"},
+        )
+        assert resp.status_code == 200
+        data = resp.json()
+        assert data["audio_url"] == "https://x.com/cloned.mp3"
+
+        # Verify CosyVoice was called with the resolved voice_id, not the UUID
+        mock_service.synthesize_speech.assert_called_once_with(
+            text="克隆音色测试",
+            voice_id="cosyvoice_actual_voice_123",
+            speed=1.0,
+        )
+        # Verify repo was queried with the UUID
+        mock_clone_repo.get.assert_called_once_with("abc123-uuid-of-profile")
+
+    def test_preview_clone_voice_incomplete_returns_400(self):
+        """Clone profile with empty voice_id returns 400."""
+        from fastapi import FastAPI
+
+        app = FastAPI()
+        from app.api.routes.tts import router
+
+        app.include_router(router, prefix="/tts")
+
+        from app.auth import get_current_user
+        from app.dependencies import get_voice_clone_profile_repository
+
+        fake_user = MagicMock()
+        fake_user.user.id = "user-1"
+        app.dependency_overrides[get_current_user] = lambda: fake_user
+
+        # Mock voice clone profile with empty voice_id (clone not finished)
+        mock_profile = MagicMock()
+        mock_profile.user_id = "user-1"
+        mock_profile.voice_id = ""
+        mock_clone_repo = MagicMock()
+        mock_clone_repo.get.return_value = mock_profile
+        app.dependency_overrides[get_voice_clone_profile_repository] = lambda: mock_clone_repo
+
+        client = self._make_client(app)
+        resp = client.post(
+            "/tts/preview",
+            json={"text": "测试未完成克隆", "voice_id": "abc123-uuid"},
+        )
+        assert resp.status_code == 400
+        assert "音色克隆尚未完成" in resp.json()["detail"]
+
+    def test_preview_preset_voice_passthrough(self):
+        """Preset voice ID (not a profile UUID) passes through unchanged."""
+        from fastapi import FastAPI
+
+        app = FastAPI()
+        from app.api.routes.tts import router
+
+        app.include_router(router, prefix="/tts")
+
+        from app.auth import get_current_user
+        from app.dependencies import get_cosyvoice_service, get_voice_clone_profile_repository
+
+        fake_user = MagicMock()
+        fake_user.user.id = "user-1"
+        app.dependency_overrides[get_current_user] = lambda: fake_user
+
+        mock_service = MagicMock()
+        mock_service.synthesize_speech.return_value = FakeSynthesizeResult(
+            audio_url="https://x.com/preset.mp3",
+            duration=2.0,
+        )
+        app.dependency_overrides[get_cosyvoice_service] = lambda: mock_service
+
+        # Mock repo returns None (preset voice, not a clone profile)
+        mock_clone_repo = MagicMock()
+        mock_clone_repo.get.return_value = None
+        app.dependency_overrides[get_voice_clone_profile_repository] = lambda: mock_clone_repo
+
+        client = self._make_client(app)
+        resp = client.post(
+            "/tts/preview",
+            json={"text": "预设音色测试", "voice_id": "longxiaoxia_v3"},
+        )
+        assert resp.status_code == 200
+
+        # Verify CosyVoice was called with the original preset voice_id
+        mock_service.synthesize_speech.assert_called_once_with(
+            text="预设音色测试",
+            voice_id="longxiaoxia_v3",
+            speed=1.0,
+        )
+
+    def test_preview_clone_voice_wrong_user_returns_403(self):
+        """Accessing another user's clone profile returns 403."""
+        from fastapi import FastAPI
+
+        app = FastAPI()
+        from app.api.routes.tts import router
+
+        app.include_router(router, prefix="/tts")
+
+        from app.auth import get_current_user
+        from app.dependencies import get_voice_clone_profile_repository
+
+        fake_user = MagicMock()
+        fake_user.user.id = "user-1"
+        app.dependency_overrides[get_current_user] = lambda: fake_user
+
+        # Mock profile belonging to a different user
+        mock_profile = MagicMock()
+        mock_profile.user_id = "user-2"
+        mock_profile.voice_id = "cosyvoice_voice_xyz"
+        mock_clone_repo = MagicMock()
+        mock_clone_repo.get.return_value = mock_profile
+        app.dependency_overrides[get_voice_clone_profile_repository] = lambda: mock_clone_repo
+
+        client = self._make_client(app)
+        resp = client.post(
+            "/tts/preview",
+            json={"text": "越权测试", "voice_id": "other-user-profile-uuid"},
+        )
+        assert resp.status_code == 403
+        assert "无权访问该音色" in resp.json()["detail"]