Update app.py

app.py

@@ -9,12 +9,14 @@ from tortoise.api import TextToSpeech
 from tortoise.utils.text import split_and_recombine_text
 from tortoise.utils.audio import load_audio, load_voice, load_voices
 
+tts = TextToSpeech(kv_cache=True)
+
 @spaces.GPU
 def inference(
     text,
     reference_audio,
-    seed,
 ):
+    output_wav_path = tempfile.mktemp(suffix=".wav")
     texts = split_and_recombine_text(text)
 
     start_time = time.time()
@@ -32,8 +34,9 @@ def inference(
 
     wav = torch.cat(all_parts, dim=0).unsqueeze(0)
     print(wav.shape)
-    torchaudio.save("output.wav", wav.cpu(), 24000)
-    yield (None, gr.make_waveform(audio="output.wav",))
+    torchaudio.save(output_wav_path, wav.cpu(), 24000)
+
+    return output_wav_path
     
 def main():
     title = "Tortoise TTS 🐢"
@@ -68,8 +71,6 @@ def main():
 
 
 if __name__ == "__main__":
-    tts = TextToSpeech(kv_cache=True, use_deepspeed=True, half=True)
-
     with open("Tortoise_TTS_Runs_Scripts.log", "a") as f:
         f.write(
             f"\n\n-------------------------Tortoise TTS Scripts Logs, {datetime.now()}-------------------------\n"