update audio output debug

api.py

@@ -4,6 +4,7 @@ import edge_tts
 import soundfile as sf
 import torch
 import fairseq
+import numpy as np
 from fastapi import FastAPI, HTTPException
 from fastapi.responses import FileResponse
 from pydantic import BaseModel
@@ -188,6 +189,12 @@ async def tts_api(req: TTSRequest):
         #             f.write(chunk["data"])
         # await generate_tts_with_back4app(req.text, req.tts_voice, tts_wav)
         await generate_tts_with_hf_space(req.text, req.tts_voice, tts_wav)
+        
+        # Validasi file audio TTS
+        if not os.path.isfile(tts_wav):
+            raise HTTPException(status_code=500, detail="TTS file was not created")
+        if os.path.getsize(tts_wav) == 0:
+            raise HTTPException(status_code=500, detail="TTS file is empty")
     except Exception as e:
         raise HTTPException(status_code=500, detail=f"TTS error: {e}")
 
@@ -201,7 +208,7 @@ async def tts_api(req: TTSRequest):
             raise Exception("Failed to load model")
 
         # Run conversion menggunakan method single() yang benar
-        result = vc.single(
+        audio_opt = vc.single(
             sid=0,                           # speaker id
             input_audio=tts_wav,         # path audio input
             embedder_model_name="auto",       # auto detect embedder
@@ -216,12 +223,21 @@ async def tts_api(req: TTSRequest):
             output_dir=output_dir            # output directory
         )
 
-        # Cek apakah result tuple atau string error
+        # Cek apakah result adalah numpy array yang valid
         print("input_audio : ", tts_wav)
-        print("result : ", result)
-        if not (isinstance(result, tuple) and isinstance(result[1], tuple)):
-            raise HTTPException(status_code=500, detail=f"RVC error: {result}")
-        info, (tgt_sr, audio_opt) = result
+        print("audio_opt type: ", type(audio_opt))
+        print("audio_opt shape: ", audio_opt.shape if hasattr(audio_opt, 'shape') else 'N/A')
+        print("audio_opt sample: ", audio_opt[:10] if hasattr(audio_opt, '__getitem__') else 'N/A')
+        
+        if not isinstance(audio_opt, np.ndarray):
+            raise HTTPException(status_code=500, detail=f"RVC error: Expected numpy array, got {type(audio_opt)}")
+        
+        # Cek apakah audio_opt tidak kosong (tidak semua zeros)
+        if len(audio_opt) == 0 or np.all(audio_opt == 0):
+            raise HTTPException(status_code=500, detail="RVC error: Generated audio is empty or all zeros. Check input audio and model configuration.")
+        
+        # Gunakan tgt_sr dari model
+        tgt_sr = vc.tgt_sr
         sf.write(output_wav, audio_opt, tgt_sr)
     except Exception as e:
         raise HTTPException(status_code=500, detail=f"RVC exception error: {e}")