Update app.py

app.py

@@ -43,26 +43,26 @@ def generate_script(user_prompt: str, model_id: str, token: str, duration: int):
         combined_prompt = f"{system_prompt}\nUser concept: {user_prompt}\nRefined script and music suggestion:"
         result = llama_pipeline(combined_prompt, max_new_tokens=500, do_sample=True, temperature=0.9)
 
-        # Check if the output contains "Music Suggestion:"
+        # Extract the script and music suggestion
         generated_text = result[0]["generated_text"]
-        if "Music Suggestion:" in generated_text:
-            script, music_suggestion = generated_text.split("Music Suggestion:", 1)
-            return script.strip(), music_suggestion.strip()
+        if "Music suggestion:" in generated_text:
+            parts = generated_text.split("Music suggestion:", 1)
+            script = parts[0].strip()
+            music_suggestion = parts[1].strip()
+            return script, music_suggestion
         else:
-            # Return the full text as the script and indicate no music suggestion was found
             return generated_text.strip(), "No specific music suggestion found."
     except Exception as e:
         return f"Error generating script: {e}", None
 
-
 # ---------------------------------------------------------------------
 # Voice-Over Generation Function
 # ---------------------------------------------------------------------
 @spaces.GPU(duration=300)
-def generate_voice(script: str, speaker: str):
+def generate_voice(script: str, speaker: str = "default"):
     try:
-        # Replace with your chosen TTS model
-        tts_model = "coqui/XTTS-v2"
+        # Replace with a real TTS model
+        tts_model = "tts_models/en/ljspeech/tacotron2-DDC"
         processor = AutoProcessor.from_pretrained(tts_model)
         model = AutoModelForCausalLM.from_pretrained(tts_model)
 
@@ -145,26 +145,38 @@ with gr.Blocks() as demo:
     blend_button = gr.Button("Blend Audio")
     final_output = gr.Audio(label="Final Promo Audio", type="filepath")
 
+    def step_generate_script(user_prompt, llama_model_id, duration):
+        return generate_script(user_prompt, llama_model_id, hf_token, duration)
+
+    def step_generate_voice(script, speaker):
+        return generate_voice(script, speaker)
+
+    def step_generate_music(music_suggestion, audio_length):
+        return generate_music(music_suggestion, audio_length)
+
+    def step_blend_audio(voice_path, music_path, ducking):
+        return blend_audio(voice_path, music_path, ducking)
+
     generate_script_button.click(
-        fn=lambda user_prompt, llama_model_id, duration: generate_script(user_prompt, llama_model_id, hf_token, duration),
+        fn=step_generate_script,
         inputs=[user_prompt, llama_model_id, duration],
         outputs=[script_output],
     )
 
     generate_voice_button.click(
-        fn=generate_voice,
+        fn=step_generate_voice,
         inputs=[script_output, speaker],
         outputs=[voice_output],
     )
 
     generate_music_button.click(
-        fn=generate_music,
+        fn=step_generate_music,
         inputs=[script_output, audio_length],
         outputs=[music_output],
     )
 
     blend_button.click(
-        fn=blend_audio,
+        fn=step_blend_audio,
         inputs=[voice_output, music_output, ducking],
         outputs=[final_output],
     )