Update app.py

app.py

@@ -2,71 +2,60 @@ import gradio as gr
 import json
 from transformers import pipeline
 from PIL import Image
-import numpy as np
 import os
 import spaces
 
-# Text Expansion Model (ZeroGPU compatible)
+# Text Expansion Model (ZeroGPU-compatible)
 text_generator = pipeline("text2text-generation", model="t5-large")
 
-# Text-to-Speech Model (ZeroGPU compatible)
-tts_pipeline = pipeline("text-to-speech", model="facebook/speecht5_tts")
-
-# Image Generation Model (LoRA-based)
+# Image Generation Model (ZeroGPU-compatible)
 image_generator = pipeline("text-to-image", model="stabilityai/sdxl-turbo")
 
-# Main Processing Function
 @spaces.GPU
 def generate_cartoon(script_text):
-    # Step 1: Expand Script
+    # Step 1: Generate scene breakdown
     prompt = f"""
-    Convert the following cartoon story into a detailed scene plan in JSON format. For each scene, provide:
-    1. Scene description
-    2. Dialogue
-    3. Characters involved
-    4. Background description
-
-    Story:
-    {script_text}
+    Convert this cartoon story into a structured scene list in JSON format. For each scene include:
+    1. "scene_description"
+    2. "dialogue"
+    3. "characters"
+    4. "background_description"
+
+    Story: {script_text}
+    Return as: {{ "scenes": [{{ ... }}] }}
     """
 
-    response = text_generator(prompt, max_new_tokens=1024)[0]['generated_text']
-
     try:
-        scene_data = json.loads(response)
-    except:
-        return "Script expansion failed. Please refine input."
+        result = text_generator(prompt, max_new_tokens=1024)[0]['generated_text']
+        scene_data = json.loads(result)
+    except Exception as e:
+        return f"Failed to generate structured scenes. Error: {e}"
 
     os.makedirs("generated_images", exist_ok=True)
-    os.makedirs("generated_audio", exist_ok=True)
-
-    scene_results = []
 
+    results = []
     for idx, scene in enumerate(scene_data.get("scenes", [])):
-        # Generate Background Image
-        background_prompt = scene.get("background_description", "cartoon background")
-        background_image = image_generator(background_prompt).images[0]
-        bg_path = f"generated_images/scene_{idx+1}.png"
-        background_image.save(bg_path)
+        # Generate background image
+        bg_prompt = scene.get("background_description", "cartoon background, colorful")
+        image = image_generator(bg_prompt, guidance_scale=7.5, num_inference_steps=25).images[0]
+        image_path = f"generated_images/scene_{idx+1}.png"
+        image.save(image_path)
 
-        # Generate TTS Audio
+        # Collect image and dialogue
         dialogue = scene.get("dialogue", "")
-        audio_output = tts_pipeline(dialogue)
-        audio_path = f"generated_audio/scene_{idx+1}.wav"
-        audio_output['audio'].export(audio_path, format="wav")
+        results.append((image_path, f"Scene {idx+1} Dialogue:\n{dialogue}"))
 
-        scene_results.append((bg_path, audio_path))
+    return results
 
-    return scene_results
 
-# Gradio UI
+# Gradio Interface
 demo = gr.Interface(
     fn=generate_cartoon,
-    inputs=gr.Textbox(label="Enter Cartoon Script", lines=10),
-    outputs=gr.Gallery(label="Generated Scenes with Audio"),
-    title="Cartoon Scene Generator",
-    description="Enter a cartoon story script to generate scenes with background images and audio. Video merging should be done separately in Google Colab."
+    inputs=gr.Textbox(label="Enter Cartoon Script", lines=10, placeholder="Once upon a time..."),
+    outputs=gr.Gallery(label="Generated Scenes with Dialogues", columns=1, type="auto"),
+    title="Cartoon Film Generator (ZeroGPU)",
+    description="Enter a cartoon story. This tool will break it into scenes, generate background images, and show dialogues. Audio and video creation should be done in Colab."
 )
 
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch()