Update app.py

app.py

@@ -295,8 +295,8 @@ def phonemes_to_ids(config, phonemes: List[str]) -> List[int]:
         ids.extend(id_map[PAD])
     ids.extend(id_map[EOS])
     return ids
-
-def inferencing(model, config, sid, line, length_scale = 1, noise_scale = 0.667, noise_scale_w = 0.8, auto_play=True):
+@app.get("/generate_audio")
+async def inferencing(model, config, sid, line, length_scale = 1, noise_scale = 0.667, noise_scale_w = 0.8, auto_play=True):
     audios = []
     if config["phoneme_type"] == "PhonemeType.ESPEAK":
         config["phoneme_type"] = "espeak"
@@ -330,8 +330,14 @@ def inferencing(model, config, sid, line, length_scale = 1, noise_scale = 0.667,
         audios.append(audio)
     merged_audio = np.concatenate(audios)
     sample_rate = config["audio"]["sample_rate"]
-    display(Markdown(f"{line}"))
-    display(Audio(merged_audio, rate=sample_rate, autoplay=auto_play))
+    # Save the audio as a temporary WAV file
+    temp_audio_path = os.path.join(tempfile.gettempdir(), "generated_audio.wav")
+    sf.write(temp_audio_path, merged_audio, config["audio"]["sample_rate"])
+    
+    # Return the audio file as a FastAPI response
+    return FileResponse(temp_audio_path)
+  #  display(Markdown(f"{line}"))
+   # display(Audio(merged_audio, rate=sample_rate, autoplay=auto_play))
 
 def denoise(
     audio: np.ndarray, bias_spec: np.ndarray, denoiser_strength: float