diff --git a/apps/api/app/api/routes/tts.py b/apps/api/app/api/routes/tts.py index eb74903ff..eb9581e07 100755 --- a/apps/api/app/api/routes/tts.py +++ b/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: diff --git a/tests/unit/test_tts_preview.py b/tests/unit/test_tts_preview.py index 4320b4938..6ff37a799 100644 --- a/tests/unit/test_tts_preview.py +++ b/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"]