feat: Implement segment update and regeneration features in Audiobook API and frontend

qwen3-tts-backend/api/audiobook.py

@@ -20,6 +20,7 @@ from schemas.audiobook import (
     AudiobookChapterResponse,
     AudiobookCharacterEdit,
     AudiobookSegmentResponse,
+    AudiobookSegmentUpdate,
     AudiobookGenerateRequest,
     AudiobookAnalyzeRequest,
 )
@@ -543,6 +544,89 @@ async def get_segments(
     return result
 
 
+@router.put("/projects/{project_id}/segments/{segment_id}", response_model=AudiobookSegmentResponse)
+async def update_segment(
+    project_id: int,
+    segment_id: int,
+    data: AudiobookSegmentUpdate,
+    current_user: User = Depends(get_current_user),
+    db: Session = Depends(get_db),
+):
+    project = crud.get_audiobook_project(db, project_id, current_user.id)
+    if not project:
+        raise HTTPException(status_code=404, detail="Project not found")
+
+    seg = db.query(AudiobookSegment).filter(
+        AudiobookSegment.id == segment_id,
+        AudiobookSegment.project_id == project_id,
+    ).first()
+    if not seg:
+        raise HTTPException(status_code=404, detail="Segment not found")
+
+    seg = crud.update_audiobook_segment(db, segment_id, data.text, data.emo_text, data.emo_alpha)
+    char_name = seg.character.name if seg.character else None
+    return AudiobookSegmentResponse(
+        id=seg.id,
+        project_id=seg.project_id,
+        chapter_index=seg.chapter_index,
+        segment_index=seg.segment_index,
+        character_id=seg.character_id,
+        character_name=char_name,
+        text=seg.text,
+        emo_text=seg.emo_text,
+        emo_alpha=seg.emo_alpha,
+        audio_path=seg.audio_path,
+        status=seg.status,
+    )
+
+
+@router.post("/projects/{project_id}/segments/{segment_id}/regenerate", response_model=AudiobookSegmentResponse)
+async def regenerate_segment(
+    project_id: int,
+    segment_id: int,
+    current_user: User = Depends(get_current_user),
+    db: Session = Depends(get_db),
+):
+    project = crud.get_audiobook_project(db, project_id, current_user.id)
+    if not project:
+        raise HTTPException(status_code=404, detail="Project not found")
+
+    seg = db.query(AudiobookSegment).filter(
+        AudiobookSegment.id == segment_id,
+        AudiobookSegment.project_id == project_id,
+    ).first()
+    if not seg:
+        raise HTTPException(status_code=404, detail="Segment not found")
+
+    from core.audiobook_service import generate_single_segment
+    from core.database import SessionLocal
+
+    async def run():
+        async_db = SessionLocal()
+        try:
+            db_user = crud.get_user_by_id(async_db, current_user.id)
+            await generate_single_segment(segment_id, db_user, async_db)
+        finally:
+            async_db.close()
+
+    asyncio.create_task(run())
+
+    char_name = seg.character.name if seg.character else None
+    return AudiobookSegmentResponse(
+        id=seg.id,
+        project_id=seg.project_id,
+        chapter_index=seg.chapter_index,
+        segment_index=seg.segment_index,
+        character_id=seg.character_id,
+        character_name=char_name,
+        text=seg.text,
+        emo_text=seg.emo_text,
+        emo_alpha=seg.emo_alpha,
+        audio_path=seg.audio_path,
+        status="generating",
+    )
+
+
 @router.get("/projects/{project_id}/segments/{segment_id}/audio")
 async def get_segment_audio(
     project_id: int,

qwen3-tts-backend/core/audiobook_service.py

@@ -672,6 +672,118 @@ async def generate_project(project_id: int, user: User, db: Session, chapter_ind
             crud.update_audiobook_project_status(db, project_id, "error", error_message=str(e))
 
 
+async def generate_single_segment(segment_id: int, user: User, db: Session) -> None:
+    from db.models import AudiobookSegment as _Seg
+    seg = db.query(_Seg).filter(_Seg.id == segment_id).first()
+    if not seg:
+        return
+
+    output_base = Path(settings.OUTPUT_DIR) / "audiobook" / str(seg.project_id) / "segments"
+    output_base.mkdir(parents=True, exist_ok=True)
+
+    crud.update_audiobook_segment_status(db, segment_id, "generating")
+    try:
+        from core.tts_service import TTSServiceFactory
+        from core.security import decrypt_api_key
+
+        backend_type = user.user_preferences.get("default_backend", "aliyun") if user.user_preferences else "aliyun"
+        user_api_key = None
+        if backend_type == "aliyun" and user.aliyun_api_key:
+            user_api_key = decrypt_api_key(user.aliyun_api_key)
+
+        backend = await TTSServiceFactory.get_backend(backend_type, user_api_key)
+
+        char = crud.get_audiobook_character(db, seg.character_id)
+        if not char or not char.voice_design_id:
+            crud.update_audiobook_segment_status(db, segment_id, "error")
+            return
+
+        design = crud.get_voice_design(db, char.voice_design_id, user.id)
+        if not design:
+            crud.update_audiobook_segment_status(db, segment_id, "error")
+            return
+
+        await _bootstrap_character_voices([seg], user, backend, backend_type, db)
+        db.refresh(design)
+
+        audio_filename = f"ch{seg.chapter_index:03d}_seg{seg.segment_index:04d}.wav"
+        audio_path = output_base / audio_filename
+
+        if backend_type == "aliyun":
+            if design.aliyun_voice_id:
+                audio_bytes, _ = await backend.generate_voice_design(
+                    {"text": seg.text, "language": "zh"},
+                    saved_voice_id=design.aliyun_voice_id
+                )
+            else:
+                audio_bytes, _ = await backend.generate_voice_design({
+                    "text": seg.text,
+                    "language": "zh",
+                    "instruct": _get_gendered_instruct(char.gender, design.instruct),
+                })
+        elif char.use_indextts2 and design.ref_audio_path and Path(design.ref_audio_path).exists():
+            from core.tts_service import IndexTTS2Backend
+            indextts2 = IndexTTS2Backend()
+            audio_bytes = await indextts2.generate(
+                text=seg.text,
+                spk_audio_prompt=design.ref_audio_path,
+                output_path=str(audio_path),
+                emo_text=seg.emo_text or None,
+                emo_alpha=seg.emo_alpha if seg.emo_text else 0.5,
+            )
+        else:
+            if design.voice_cache_id:
+                from core.cache_manager import VoiceCacheManager
+                cache_manager = await VoiceCacheManager.get_instance()
+                cache_result = await cache_manager.get_cache_by_id(design.voice_cache_id, db)
+                x_vector = cache_result['data'] if cache_result else None
+                if x_vector:
+                    audio_bytes, _ = await backend.generate_voice_clone(
+                        {
+                            "text": seg.text,
+                            "language": "Auto",
+                            "max_new_tokens": 2048,
+                            "temperature": 0.3,
+                            "top_k": 10,
+                            "top_p": 0.9,
+                            "repetition_penalty": 1.05,
+                        },
+                        x_vector=x_vector
+                    )
+                else:
+                    audio_bytes, _ = await backend.generate_voice_design({
+                        "text": seg.text,
+                        "language": "Auto",
+                        "instruct": _get_gendered_instruct(char.gender, design.instruct),
+                        "max_new_tokens": 2048,
+                        "temperature": 0.3,
+                        "top_k": 10,
+                        "top_p": 0.9,
+                        "repetition_penalty": 1.05,
+                    })
+            else:
+                audio_bytes, _ = await backend.generate_voice_design({
+                    "text": seg.text,
+                    "language": "Auto",
+                    "instruct": _get_gendered_instruct(char.gender, design.instruct),
+                    "max_new_tokens": 2048,
+                    "temperature": 0.3,
+                    "top_k": 10,
+                    "top_p": 0.9,
+                    "repetition_penalty": 1.05,
+                })
+
+        with open(audio_path, "wb") as f:
+            f.write(audio_bytes)
+
+        crud.update_audiobook_segment_status(db, segment_id, "done", audio_path=str(audio_path))
+        logger.info(f"Single segment {segment_id} generated: {audio_path}")
+
+    except Exception as e:
+        logger.error(f"Single segment {segment_id} generation failed: {e}", exc_info=True)
+        crud.update_audiobook_segment_status(db, segment_id, "error")
+
+
 def merge_audio_files(audio_paths: list[str], output_path: str) -> None:
     from pydub import AudioSegment
 

qwen3-tts-backend/core/llm_service.py

@@ -131,13 +131,12 @@ class LLMService:
             "5. 性格特质：核心性格、情绪模式、表达习惯\n"
             "6. 叙事风格：语速节奏、停顿习惯、语气色彩、整体叙述感\n\n"
             "注意：instruct 的第一行（音色信息）必须与 gender 字段保持一致。如果 gender 为女，第一行绝对不能出现'男性'字样。\n\n"
-            "【特别规定】narrator（旁白）的 instruct 必须固定描述为传统说书人风格，参考如下模板（根据书籍风格可微调措辞，但风格不变）：\n"
-            "音色信息：浑厚醇厚的男性中低音，嗓音饱满有力，带有传统说书人的磁性与感染力\n"
-            "身份背景：中国传统说书艺人，精通评书、章回小说叙述艺术，深谙故事节奏与听众心理\n"
-            "年龄设定：中年男性，四五十岁，声音历经岁月沉淀，成熟稳重而不失活力\n"
-            "外貌特征：面容沉稳，气度从容，台风大气，给人以可信赖的叙述者印象\n"
-            "性格特质：沉稳睿智，叙事冷静客观，情到深处能引发共鸣，不动声色间娓娓道来\n"
-            "叙事风格：语速适中偏慢，抑扬顿挫，擅长铺垫悬念，停顿恰到好处，语气庄重而生动，富有画面感\n\n"
+            "【特别规定】narrator（旁白）的 instruct 必须根据小说类型选择对应的叙述者音色风格，规则如下：\n"
+            "▸ 古风/武侠/历史/玄幻/仙侠/奇幻 → 传统说书人风格：浑厚醇厚的男性中低音，嗓音饱满有力，带有说书人的磁性与感染力；中年男性，四五十岁；语速适中偏慢，抑扬顿挫，停顿恰到好处，语气庄重生动，富有画面感\n"
+            "▸ 现代言情/都市爱情/青春校园 → 年轻女性叙述者风格：女性声音，清亮柔和的中高音，嗓音清新干净，带有亲切温柔的娓娓道来感；二三十岁年轻女性；语速轻快自然，情感细腻，语气温柔而富有感染力\n"
+            "▸ 悬疑/推理/惊悚/恐怖 → 低沉神秘风格：男性声音，低沉压抑的男性低音，嗓音干练克制，带有一丝神秘与张力；中年男性；语速沉稳偏慢，停顿制造悬念，语气冷静克制，暗藏紧张感\n"
+            "▸ 科幻/末世/赛博朋克 → 理性宏观风格：男性声音，清晰有力的男性中音，嗓音冷静客观，带有纪录片解说员的宏大叙事感；语速稳定，条理清晰，语气客观宏观，富有科技感与史诗感\n"
+            "▸ 其他/无法判断 → 传统说书人风格（同古风类型）\n\n"
             "只输出JSON，格式如下，不要有其他文字：\n"
             '{"characters": [{"name": "narrator", "gender": "未知", "description": "第三人称叙述者", "instruct": "音色信息：...\\n身份背景：...\\n年龄设定：...\\n外貌特征：...\\n性格特质：...\\n叙事风格：..."}, ...]}'
         )

qwen3-tts-backend/db/crud.py

@@ -640,6 +640,26 @@ def update_audiobook_segment_status(
     return seg
 
 
+def update_audiobook_segment(
+    db: Session,
+    segment_id: int,
+    text: str,
+    emo_text: Optional[str],
+    emo_alpha: Optional[float],
+) -> Optional[AudiobookSegment]:
+    seg = db.query(AudiobookSegment).filter(AudiobookSegment.id == segment_id).first()
+    if not seg:
+        return None
+    seg.text = text
+    seg.emo_text = emo_text or None
+    seg.emo_alpha = emo_alpha
+    seg.status = "pending"
+    seg.audio_path = None
+    db.commit()
+    db.refresh(seg)
+    return seg
+
+
 def delete_audiobook_segments(db: Session, project_id: int) -> None:
     db.query(AudiobookSegment).filter(AudiobookSegment.project_id == project_id).delete()
     db.commit()

qwen3-tts-backend/schemas/audiobook.py

@@ -89,6 +89,12 @@ class AudiobookSegmentResponse(BaseModel):
     model_config = ConfigDict(from_attributes=True)
 
 
+class AudiobookSegmentUpdate(BaseModel):
+    text: str
+    emo_text: Optional[str] = None
+    emo_alpha: Optional[float] = None
+
+
 class LLMConfigUpdate(BaseModel):
     base_url: str
     api_key: str