add edge tts

scheduler/jobs.py

@@ -3,6 +3,8 @@ from llm import prompt
 from utils.logger import logger
 import datetime
 from llm.generate_podcast import generate_topics
+import os
+import asyncio
 from models.script import Script
 from config.database import SessionLocal
 
@@ -152,8 +154,110 @@ def job_generate_script():
         db.rollback()
         logger.error(f"Failed to generate/save full script: {e}")
 
+
+def job_synthesize_podcast_audio():
+    """定时任务：从数据库读取最新完整脚本，调用 TTS 服务生成整期播客音频并保存到磁盘。"""
+    logger.info("Starting job_synthesize_podcast_audio")
+    db = SessionLocal()
+    try:
+        script = db.query(Script).filter_by(project="梗文化研究所").order_by(Script.create_time.desc()).first()
+        if not script or not script.content:
+            logger.warning("No script found for synthesizing podcast audio.")
+            return
+
+        data = json.loads(script.content)
+        full_script = data.get("script") or {}
+        if not full_script:
+            logger.warning("No 'script' section found in latest Script record.")
+            return
+
+        title = full_script.get("title") or f"podcast_{script.id}"
+        script_items = full_script.get("script", [])
+        if not script_items:
+            logger.warning("Empty script items, nothing to synthesize.")
+            return
+
+        # 按段落分别合成（使用 TTSService 的 text/voice 参数），再拼接音频
+        from config.settings import settings
+        from tts.service import TTSService
+
+        # 角色到声音的映射（可按需扩展或放到配置中）
+        role_voice_map = {
+            "host": settings.TTS_VOICE or "zh-CN-XiaoxiaoNeural",
+            "guest": "zh-CN-YunxiNeural",
+            # fallback for other roles
+            "default": settings.TTS_VOICE or "zh-CN-XiaoxiaoNeural",
+        }
+
+        segment_audio_bytes = []
+
+        for idx, item in enumerate(script_items):
+            role = (item.get("role") or "").lower()
+            text = item.get("text", "").strip()
+            if not text:
+                continue
+
+            voice = role_voice_map.get(role, role_voice_map["default"]) if role else role_voice_map["default"]
+
+            try:
+                logger.debug(f"Synthesizing segment {idx} role={role} voice={voice} text='{text[:30]}...'")
+                seg_audio = asyncio.run(TTSService.synthesize(text=text, voice=voice, language=settings.TTS_LANGUAGE))
+                segment_audio_bytes.append((idx, role or "segment", seg_audio))
+                logger.debug(f"Synthesized segment {idx} role={role} size={seg_audio.getbuffer().nbytes}")
+            except Exception as e:
+                logger.error(f"Failed to synthesize segment {idx} (role={role}): {e}")
+
+        if not segment_audio_bytes:
+            logger.warning("No audio segments synthesized; aborting podcast save.")
+            return
+
+        # 保存或合并音频：优先使用 pydub (ffmpeg)，否则保存为独立段文件
+        out_dir = os.path.join("output", "podcasts")
+        os.makedirs(out_dir, exist_ok=True)
+        safe_title = "_".join(title.split())
+        final_filename = f"{safe_title}_{script.subject}_{script.id}.mp3"
+        final_path = os.path.join(out_dir, final_filename)
+
+        try:
+            from pydub import AudioSegment
+
+            combined = None
+            for idx, role, seg in sorted(segment_audio_bytes, key=lambda x: x[0]):
+                seg.seek(0)
+                audio_seg = AudioSegment.from_file(seg, format="mp3")
+                if combined is None:
+                    combined = audio_seg
+                else:
+                    combined = combined + audio_seg
+
+            if combined is not None:
+                combined.export(final_path, format="mp3")
+                logger.info(f"Saved combined podcast audio to {final_path}")
+                return
+
+        except Exception as e:
+            logger.warning(f"pydub/ffmpeg not available or merge failed: {e}; falling back to per-segment files")
+
+        # 回退：保存每个分段为独立文件，并记录它们
+        segment_paths = []
+        for idx, role, seg in sorted(segment_audio_bytes, key=lambda x: x[0]):
+            seg.seek(0)
+            seg_filename = f"{safe_title}_{script.subject}_{script.id}_seg{idx}_{role}.mp3"
+            seg_path = os.path.join(out_dir, seg_filename)
+            with open(seg_path, "wb") as fw:
+                fw.write(seg.getvalue())
+            segment_paths.append(seg_path)
+
+        logger.info(f"Saved {len(segment_paths)} segment files to {out_dir}; combined file not created")
+
+    except Exception as e:
+        logger.error(f"Failed to synthesize/save podcast audio: {e}")
+    finally:
+        db.close()
+
 # For manual testing
 if __name__ == "__main__":
     # job_generate_topics()
     # job_generate_bits()
-    job_generate_script()
+    # job_generate_script()
+    job_synthesize_podcast_audio()