feat(#1898): TTS 参数适配 CosyVoice v3 官方 API

CosyVoice v3 API 调整：emotion 字段已废弃，改用 input.instruction 中文自然语言指令；语言通过 input.language_hints 数组传递（仅取第一个元素）。 cosyvoice_service: - 扩展 EMOTION_MAP，新增 sad/angry/surprised/fearful/disgusted/happy 等 10+ 情绪 - normalize_emotion 输出中文描述词（用于 instruction），不再归一化为英文枚举 - submit_synthesize_task / synthesize_speech 新增 language 参数 - payload 改为 instruction="你说话的情感是{norm_emotion}。" + language_hints=[lang] - 系统音色仅传 zh/en；克隆音色不做语言限制（v3 克隆音色支持多语言） - rate 字段名保持（CosyVoice 官方文档确认仍用 rate 字段） schemas/routes: - TTS 预览请求新增 language 字段（默认 zh-CN） - emotion 字段描述更新为支持中文/英文自然语言指令 - /tts/synthesize 的 synthesis_meta 透传 language - lipsync 任务默认 language=zh tts_job workflow / streaming_service: - 所有 CosyVoice 调用点透传 language 参数（默认 zh-CN） - metadata 中 language 字段从请求一路透传到分段合成线程池新增单测: - test_normalize_emotion_* 更新为中文描述词断言，新增情绪覆盖 - test_submit_synthesize_payload_uses_instruction_and_language_hints - test_submit_synthesize_payload_english_emotion_maps_to_chinese - 相关老测试补 language kwarg 断言 - 全量 15127 passed，0 failed [skip ci-format-check]
2026-09-15 04:10:18 +08:00
parent 695a491c5d
commit 38e4dfd628
12 changed files with 156 additions and 48 deletions
@@ -177,6 +177,7 @@ def synthesize(
    synthesis_meta = {
        "speed": request.speed,
        "emotion": request.emotion or "",
+        "language": request.language or "zh-CN",
    }
    if request.metadata_:
        synthesis_meta.update(request.metadata_)
@@ -462,10 +463,17 @@ def save_tts_job_to_library(
        try:
            proc = subprocess.run(
                [
-                    "ffprobe", "-v", "quiet", "-print_format", "json",
-                    "-show_format", str(tmp_path),
+                    "ffprobe",
+                    "-v",
+                    "quiet",
+                    "-print_format",
+                    "json",
+                    "-show_format",
+                    str(tmp_path),
                ],
-                capture_output=True, text=True, timeout=10,
+                capture_output=True,
+                text=True,
+                timeout=10,
            )
            if proc.returncode == 0:
                fmt = json.loads(proc.stdout).get("format", {})
@@ -576,6 +584,7 @@ def preview_tts(
            voice_id=actual_voice_id,
            speed=request.speed,
            emotion=request.emotion,
+            language=getattr(request, "language", "zh-CN"),
        )
    except CosyVoiceError as e:
        raise HTTPException(
@@ -67,7 +67,7 @@ class CreateLipsyncJobRequest(BaseModel):
    voice_id: str = Field("", description="音色 ID（预置音色或克隆音色 profile UUID）")
    script_text: str = Field("", description="要合成的文案（直生模式必填，最长 5000 字符）")
    speed: float = Field(1.0, ge=0.5, le=2.0, description="语速（0.5-2.0），默认 1.0")
-    emotion: str = Field("", description="情绪（natural/excited/calm/friendly 或中文 自然/兴奋/沉稳/亲切）")
+    emotion: str = Field("", description="情绪（中文/英文：自然/兴奋/沉稳/亲切/开心/悲伤/愤怒/惊讶/恐惧/厌恶 等）")

    enable_video_loop: bool = Field(
        True, description="音频长于视频时是否循环画面（AI数字人默认开启，防止音频长于视频被截断）"
@@ -16,7 +16,10 @@ class TTSSynthesizeRequest(BaseModel):
    output_name: str = Field("", description="输出文件名")
    language: str = Field("zh-CN", description="语言")
    speed: float = Field(1.0, ge=0.5, le=2.0, description="语速")
-    emotion: str = Field("", description="情绪（natural/excited/calm/friendly，或中文 自然/兴奋/沉稳/亲切）")
+    emotion: str = Field(
+        "",
+        description="情绪（中文/英文：自然/兴奋/沉稳/亲切/开心/悲伤/愤怒/惊讶/恐惧/厌恶 等；通过 instruction 自然语言指令控制）",
+    )
    voice_model: str = Field("", description="语音模型名称")
    voice_clone_profile_id: str = Field("", description="关联的音色克隆档案 ID")
    format: str = Field("mp3", description="输出格式（mp3/wav/pcm）")
@@ -110,7 +113,8 @@ class TTSPreviewRequest(BaseModel):
    text: str = Field(..., min_length=1, max_length=200, description="合成文本，限制 200 字")
    voice_id: str = Field(..., min_length=1, description="音色 ID")
    speed: float = Field(1.0, ge=0.5, le=2.0, description="语速")
-    emotion: str = Field("", description="情绪（natural/excited/calm/friendly，或中文）")
+    emotion: str = Field("", description="情绪（中文/英文：自然/兴奋/沉稳/亲切/开心/悲伤/愤怒/惊讶/恐惧/厌恶 等）")
+    language: str = Field("zh-CN", description="语言（zh-CN/en-US 等）")
    pitch: float = Field(1.0, ge=0.5, le=2.0, description="音调（预留，当前未使用）")


@@ -35,7 +35,7 @@ from app.tasks.lipsync_tts import tts_synthesize_and_submit
 from sqlalchemy.orm import Session

 from packages.adapters.sqlalchemy_impl.models import LipsyncJobModel
-from packages.application.cosyvoice_service import CosyVoiceError, normalize_emotion
+from packages.application.cosyvoice_service import CosyVoiceError
 from packages.domain.sentence_timings import (
    compute_sentence_timings,
    probe_audio_duration,
@@ -121,7 +121,8 @@ class LipsyncService:
                text=script_text,
                voice_id=actual_voice_id,
                speed=speed,
-                emotion=normalize_emotion(emotion),
+                emotion=emotion,  # normalize 在 CosyVoiceService 内部完成
+                language="zh",
            )
        except CosyVoiceError as exc:
            raise MediaKitError(f"TTS 合成失败: {exc}", code="TTSSynthesisFailed") from exc
@@ -304,7 +305,7 @@ class LipsyncService:
            voice_id=voice_id or "",
            script_text=script_text or "",
            speed=speed,
-            emotion=normalize_emotion(emotion) if is_tts_mode else (emotion or ""),
+            emotion=emotion or "",
            # 音频直传（含预合成）直接进入 pending（后续同步改为 submitted）；TTS 模式进入 tts_processing
            status="tts_processing" if is_tts_mode else "pending",
        )
@@ -325,7 +326,7 @@ class LipsyncService:
                        voice_id,
                        script_text,
                        speed,
-                        normalize_emotion(emotion),
+                        emotion or "",
                    )
                )
            except Exception as exc:
@@ -382,7 +383,8 @@ class LipsyncService:
                text=script_text,
                voice_id=actual_voice_id,
                speed=speed,
-                emotion=normalize_emotion(emotion),
+                emotion=emotion,  # normalize 在 CosyVoiceService 内部完成
+                language="zh",
            )
        except CosyVoiceError as exc:
            raise MediaKitError(f"TTS 合成失败: {exc}", code="TTSSynthesisFailed") from exc
@@ -160,6 +160,7 @@ def tts_synthesize_and_submit(
                voice_id=voice_id,
                speed=speed,
                emotion=emotion,
+                language="zh",
            )
        except CosyVoiceError as exc:
            logger.error("[lipsync_tts] TTS 合成失败: job_id=%s err=%s", job_id, exc)