Update app.py

app.py

@@ -80,13 +80,6 @@ text_client = InferenceClient(
     "mistralai/Mistral-7B-Instruct-v0.1"
 )
 
-###### COQUI TTS FUNCTIONS ######
-def get_latents(speaker_wav):
-    # create as function as we can populate here with voice cleanup/filtering
-    gpt_cond_latent, diffusion_conditioning, speaker_embedding = model.get_conditioning_latents(audio_path=speaker_wav)
-    return gpt_cond_latent, diffusion_conditioning, speaker_embedding
-
-
 def format_prompt(message, history):
   prompt = "<s>"
   for user_prompt, bot_response in history:
@@ -185,6 +178,8 @@ def bot(history, system_prompt=""):
         yield history  
 
 
+
+########### COQUI TTS FUNCTIONS #############
 def get_latents(speaker_wav):
     # Generate speaker embedding and latents for TTS
     gpt_cond_latent, diffusion_conditioning, speaker_embedding = model.get_conditioning_latents(audio_path=speaker_wav)
@@ -257,6 +252,7 @@ def generate_speech(history):
             else:
                 print("RuntimeError: non device-side assert error:", str(e))
                 raise e
+    
     #Spoken on autoplay everysencen now produce a concataned one at the one
     #requires pip install ffmpeg-python
     files_to_concat= [ffmpeg.input(w) for w in wav_list]