Update app.py

app.py

@@ -43,9 +43,14 @@ 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=200, do_sample=True, temperature=0.9)
 
-        generated_text = result[0]["generated_text"].split("Refined script and music suggestion:")[-1].strip()
-        script, music_suggestion = generated_text.split("Music Suggestion:")
-        return script.strip(), music_suggestion.strip()
+        generated_text = result[0]["generated_text"]
+
+        # Check if the delimiter exists
+        if "Music Suggestion:" in generated_text:
+            script, music_suggestion = generated_text.split("Music Suggestion:", 1)
+            return script.strip(), music_suggestion.strip()
+        else:
+            return "Error: 'Music Suggestion:' delimiter not found in the generated text.", None
     except Exception as e:
         return f"Error generating script: {e}", None
 
@@ -55,16 +60,12 @@ def generate_script(user_prompt: str, model_id: str, token: str, duration: int):
 @spaces.GPU(duration=300)
 def generate_voice(script: str, speaker: str):
     try:
-        # Replace with your chosen TTS model
-        tts_model = "coqui/XTTS-v2"
-        processor = AutoProcessor.from_pretrained(tts_model)
-        model = AutoModelForCausalLM.from_pretrained(tts_model)
+        from TTS.api import TTS
 
-        inputs = processor(script, return_tensors="pt")
-        speech = model.generate(**inputs)
+        tts = TTS("tts_models/multilingual/multi-dataset/xtts_v2", gpu=True)
 
         output_path = f"{tempfile.gettempdir()}/generated_voice.wav"
-        write(output_path, 22050, speech.cpu().numpy())
+        tts.tts_to_file(text=script, file_path=output_path)
         return output_path
     except Exception as e:
         return f"Error generating voice-over: {e}"
@@ -139,38 +140,26 @@ 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=step_generate_script,
-        inputs=[user_prompt, llama_model_id, duration],
+        fn=generate_script,
+        inputs=[user_prompt, llama_model_id, hf_token, duration],
         outputs=[script_output],
     )
 
     generate_voice_button.click(
-        fn=step_generate_voice,
+        fn=generate_voice,
         inputs=[script_output, speaker],
         outputs=[voice_output],
     )
 
     generate_music_button.click(
-        fn=step_generate_music,
+        fn=generate_music,
         inputs=[script_output, audio_length],
         outputs=[music_output],
     )
 
     blend_button.click(
-        fn=step_blend_audio,
+        fn=blend_audio,
         inputs=[voice_output, music_output, ducking],
         outputs=[final_output],
     )