Upload app.py with huggingface_hub

app.py

@@ -53,8 +53,8 @@ def openai_chat_completion(prompt, model="gpt-3.5-turbo", temperature=0.8, max_t
         st.error(f"Error con la API de OpenAI: {str(e)}")
         return None
 
-def openai_generate_image(prompt):
-    """Genera imágenes con DALL-E 3 directamente a través de requests"""
+def openai_generate_image(prompt, size="1024x1024", quality="standard"):
+    """Generate images with DALL-E 3 directly via requests"""
     headers = {
         "Content-Type": "application/json",
         "Authorization": f"Bearer {openai_api_key}"
@@ -64,28 +64,28 @@ def openai_generate_image(prompt):
         "model": "dall-e-3",
         "prompt": prompt,
         "n": 1,
-        "size": "1024x1024",
-        "quality": "standard"
+        "size": size,
+        "quality": quality
     }
     
     try:
-        with st.spinner("Creando ilustración mágica..."):
+        with st.spinner("Creating magical illustration..."):
             response = requests.post(
                 "https://api.openai.com/v1/images/generations",
                 headers=headers,
                 json=payload,
-                timeout=60
+                timeout=90  # Extended timeout for high quality images
             )
             
             if response.status_code != 200:
-                st.error("No se pudo generar la ilustración. Se usará una imagen de muestra.")
-                return "https://via.placeholder.com/512x512.png?text=Imagen+No+Disponible"
+                st.error(f"Could not generate the illustration. Using placeholder image. Error: {response.text}")
+                return "https://via.placeholder.com/512x512.png?text=Image+Not+Available"
                 
             return response.json()["data"][0]["url"]
     except Exception as e:
-        # Error discreto sin mensajes técnicos
-        st.error("Ocurrió un problema al crear la ilustración.")
-        return "https://via.placeholder.com/512x512.png?text=Imagen+No+Disponible"
+        # Discrete error without technical messages
+        st.error(f"There was a problem creating the illustration: {str(e)}")
+        return "https://via.placeholder.com/512x512.png?text=Image+Not+Available"
 
 # Optional TTS setup
 # Uncomment if using Google Cloud TTS
@@ -179,38 +179,73 @@ def split_into_scenes(text: str, num_scenes: int = 3) -> list[str]:
         return [" ".join(words[i*chunk_size : (i+1)*chunk_size]) for i in range(num_scenes)]
 
 # Image generation with DALL-E 3
-def generate_image_dalle(prompt: str, protagonist: str) -> str:
+def generate_image_dalle(prompt: str, protagonist: str, style: str = "Disney/Pixar") -> str:
     """
-    Call OpenAI Image API to create one 512×512 image from the prompt.
+    Call OpenAI Image API to create one 1024x1024 image from the prompt with distinct visual styles.
     Returns the image URL.
     """
-    # Prompt predeterminado para generación de imágenes
-    default_template = (
-        "Create a colorful children's book style illustration showing: {scene_description} " 
-        "With {protagonist_name} as the main character. Use bright colors, simple shapes, " 
-        "and a child-friendly art style."
-    )
+    # Get character gender for proper styling
+    character_gender = "gender-neutral"  # Default
+    
+    # Comprehensive style definitions with strong visual differences
+    style_templates = {
+        "Disney/Pixar": {
+            "description": "3D Pixar animation style",
+            "prompt": f"Create a stunning Pixar/Disney 3D animation style illustration. The scene shows {protagonist} {{scene_description}}. Render in the exact distinctive style of Pixar's 'Coco' or 'Soul' with volumetric lighting, subsurface scattering on skin, and detailed texturing. Use vibrant colors, expressive character designs with large eyes, and cinematic composition. The image should have depth of field, strong emotional expressions, and a polished, rendered feel."
+        },
+        
+        "Watercolor": {
+            "description": "Delicate watercolor painting",
+            "prompt": f"Create a delicate watercolor illustration of {protagonist} {{scene_description}}. Use soft transparent layers with visible paper texture and bleeding colors. The style should resemble classic Beatrix Potter or Maurice Sendak watercolors with gentle brush strokes, subtle color washes, and minimal line work. Include soft edges, color gradients, and the distinctive bleeding effect of watercolor on paper. The palette should use pastel tones with occasional vivid accents."
+        },
+        
+        "Comic Book": {
+            "description": "Bold comic book art",
+            "prompt": f"Create a dynamic comic book style illustration showing {protagonist} {{scene_description}}. Use the distinctive style of modern comic books with bold black outlines, flat color fills, dramatic perspectives, and action lines. Include comic-specific elements like dramatic shadows, exaggerated expressions, and dynamic poses. The colors should be vibrant with strong contrasts, reminiscent of Marvel or DC comic art with cel-shading techniques."
+        },
+        
+        "Claymation": {
+            "description": "Handcrafted clay animation",
+            "prompt": f"Create a claymation/stop-motion style illustration of {protagonist} {{scene_description}}. The image should look exactly like a photograph of handcrafted clay figures in a miniature set, similar to Aardman's 'Wallace and Gromit' or Laika's work. Show visible fingerprint textures in the clay, slightly imperfect proportions, and the distinctive matte finish of modeling clay. Include miniature handcrafted props, visible set construction, and the characteristic charm of stop-motion."
+        },
+        
+        "Anime": {
+            "description": "Japanese anime style",
+            "prompt": f"Create a Japanese anime-style illustration showing {protagonist} {{scene_description}}. Use the distinctive anime aesthetic with large expressive eyes, simplified facial features, and stylized colorful hair. The style should feature cel-shaded colors, speed lines for movement, and exaggerated emotional expressions. Include anime-specific visual elements like dramatic lighting effects, simplified backgrounds with depth, and the clean linework characteristic of Studio Ghibli or modern anime productions."
+        },
+        
+        "Storybook": {
+            "description": "Classic children's book illustration",
+            "prompt": f"Create a classic children's storybook illustration showing {protagonist} {{scene_description}}. The style should resemble vintage children's books with intricate hand-drawn details, rich textures, and a warm, nostalgic quality. Use the distinctive illustration style of classic illustrators like E.H. Shepard (Winnie the Pooh) or Quentin Blake (Roald Dahl books) with detailed linework, subtle watercolor washes, and charming character designs. Include fine pen details, crosshatching, and the distinctive page-like quality of traditional book illustrations."
+        }
+    }
     
-    # Intentar cargar el prompt desde archivo con fallback al predeterminado
-    img_prompt_template = load_prompt("prompts/image_prompt.txt", default_template)
+    # Get the specific style information
+    style_info = style_templates.get(style, style_templates["Disney/Pixar"])
     
-    # Format the prompt with user inputs
-    full_prompt = img_prompt_template.format(
-        scene_description=prompt,
-        protagonist_name=protagonist
-    )
+    # Base prompt with the style-specific instructions
+    base_prompt = style_info["prompt"].format(scene_description=prompt)
+    
+    # Add strong anti-text instructions
+    full_prompt = f"{base_prompt}\n\nCRITICAL REQUIREMENTS:\n" \
+                 f"1. The image MUST NOT contain ANY text, words, letters, numbers, or writing of any kind.\n" \
+                 f"2. Do not include speech bubbles, captions, labels, signs, or any other textual elements.\n" \
+                 f"3. Focus exclusively on illustrating the visual scene without attempting to include any written language.\n" \
+                 f"4. The illustration should communicate entirely through visual means only.\n\n" \
+                 f"Create a complete, finished illustration with a clear foreground and background."
     
     try:
-        # Usar nuestra función personalizada que se comunica directamente con la API
-        image_url = openai_generate_image(full_prompt)
+        # Use our custom function that communicates directly with the API
+        # Use a larger image size for better quality
+        image_url = openai_generate_image(full_prompt, size="1024x1024", quality="hd")
         if image_url:
             return image_url
         else:
-            raise Exception("No se pudo generar la imagen")
+            raise Exception("Could not generate the image")
     except Exception as e:
-        st.error(f"Error generating image: {e}")
+        st.error("Could not generate the illustration")
         # Return a placeholder image URL
-        return "https://via.placeholder.com/512x512.png?text=Image+Generation+Failed"
+        return "https://via.placeholder.com/512x512.png?text=Image+Not+Available"
 
 # Optional: Image generation with Replicate (Stable Diffusion)
 def generate_image_replicate(prompt: str, protagonist: str) -> str:
@@ -420,7 +455,7 @@ def main():
     st.title("🧐‍♂️ CuentAI – AI Story Generator")
     st.markdown("### Create personalized stories with AI-generated images")
     
-    # Sidebar with explanation
+    # Sidebar with explanation and info moved to sidebar
     with st.sidebar:
         st.subheader("About CuentAI")
         st.write("""
@@ -436,66 +471,114 @@ def main():
         
         st.subheader("Technologies")
         st.write("""
-        - OpenAI GPT-4 for generating text
+        - OpenAI GPT-3.5 for generating text
         - DALL-E 3 for creating illustrations
         - Streamlit for the web interface
         """)
         
-        # Image generation options
-        st.subheader("Options")
-        image_generator = st.radio(
-            "Image generation engine:",
-            options=["DALL-E 3", "Stable Diffusion (Replicate)"],
-            index=0
-        )
-        st.session_state.image_generator = image_generator
+        # Optional TTS toggle moved to sidebar
+        if "story" in st.session_state:
+            st.markdown("---")
+            st.subheader("🔊 Audio Narration")
+            
+            if st.checkbox("Include audio narration"):
+                # Check if TTS is imported
+                try:
+                    from google.cloud import texttospeech
+                    with st.spinner("Generating audio..."):
+                        if "audio_file" not in st.session_state:
+                            audio_file = generate_audio_tts(st.session_state.story)
+                            st.session_state.audio_file = audio_file
+                        
+                        if st.session_state.audio_file:
+                            st.audio(st.session_state.audio_file)
+                        else:
+                            st.warning("Could not generate audio. Please check your Google Cloud configuration.")
+                except ImportError:
+                    st.warning("""
+                    The narration feature requires Google Cloud Text-to-Speech.
+                    
+                    To enable this feature:
+                    1. Install the library: `pip install google-cloud-texttospeech`
+                    2. Configure your Google Cloud credentials
+                    """)
     
-    # Input form
-    col1, col2 = st.columns(2)
+    # Main content area with two columns: inputs on left, display on right
+    main_cols = st.columns([1, 1])
     
-    with col1:
+    # Left column: Input form
+    with main_cols[0]:
         st.subheader("Customize your story")
-        protagonist = st.text_input("Protagonist Name", "Alice")
-        theme = st.text_input("Story Theme", "explores a magical jungle")
-        num_scenes = st.slider("Number of scenes", min_value=1, max_value=5, value=3)
-        
-        generate_button = st.button("✨ Generate Story")
         
-        if generate_button:
-            with st.spinner("Writing story with GPT-4..."):
-                story_text = generate_story(protagonist, theme)
-                st.session_state.story = story_text
-                st.session_state.protagonist = protagonist
-                st.session_state.scenes = split_into_scenes(story_text, num_scenes=num_scenes)
-    
-    with col2:
+        # Form inputs
+        with st.container():
+            protagonist = st.text_input("Protagonist Name", "Alice")
+            theme = st.text_input("Story Theme", "exploring a magical jungle")
+            num_scenes = st.slider("Number of scenes", min_value=1, max_value=5, value=3)
+            
+            # Image generation options
+            st.subheader("Image Settings")
+            
+            # Image generator selection
+            image_generator = st.radio(
+                "Image generation engine:",
+                options=["DALL-E 3", "Stable Diffusion (Replicate)"],
+                index=0,
+                horizontal=True
+            )
+            st.session_state.image_generator = image_generator
+            
+            # Image style selection
+            image_style = st.selectbox(
+                "Illustration style:",
+                options=["Disney/Pixar", "Watercolor", "Comic Book", "Claymation", "Anime", "Storybook"],
+                index=0
+            )
+            st.session_state.image_style = image_style
+            
+            generate_button = st.button("✨ Generate Story", use_container_width=True)
+            
+            if generate_button:
+                with st.spinner("Writing story with AI..."):
+                    story_text = generate_story(protagonist, theme)
+                    st.session_state.story = story_text
+                    st.session_state.protagonist = protagonist
+                    st.session_state.scenes = split_into_scenes(story_text, num_scenes=num_scenes)
+    
+    # Right column: Display area (sample image or generated content)
+    with main_cols[1]:
         if "story" not in st.session_state:
+            # Show sample image when no story has been generated
             st.image("https://img.freepik.com/free-vector/hand-drawn-fairy-tale-castle_23-2149423879.jpg", 
                     caption="Sample image - Generate your personalized story", 
                     use_column_width=True)
-    
-    # Display story and images
-    if "story" in st.session_state:
-        st.markdown("---")
-        st.subheader("📚 Your Personalized Story")
-        
-        # Choose one layout: tabs, pagination, or scroll
-        tabs = st.tabs([f"Scene {i+1}" for i in range(len(st.session_state.scenes))])
-        
-        for i, (tab, scene) in enumerate(zip(tabs, st.session_state.scenes)):
-            with tab:
-                col1, col2 = st.columns([1, 1])
+        else:
+            # If story exists, add state to track which tab is selected
+            if "selected_tab" not in st.session_state:
+                st.session_state.selected_tab = 0
                 
-                with col1:
+            # Create tabs for scene navigation
+            tabs = st.tabs([f"Scene {i+1}" for i in range(len(st.session_state.scenes))])
+            
+            # Process each scene in tabs
+            for i, (tab, scene) in enumerate(zip(tabs, st.session_state.scenes)):
+                with tab:
+                    # When a tab is clicked, update the selected tab index
+                    st.session_state.selected_tab = i
+                    
                     # Generate image if not already in session state
                     if f"image_url_{i}" not in st.session_state:
                         with st.spinner("Generating illustration..."):
-                            # Get first 100 words for the prompt to avoid token limits
-                            scene_summary = " ".join(scene.split()[:100])
+                            # Get a meaningful summary for the image prompt
+                            scene_summary = " ".join(scene.split()[:100])  # Limit to 100 words
                             
                             # Use selected image generator
                             if st.session_state.image_generator == "DALL-E 3":
-                                img_url = generate_image_dalle(scene_summary, st.session_state.protagonist)
+                                img_url = generate_image_dalle(
+                                    scene_summary, 
+                                    st.session_state.protagonist,
+                                    st.session_state.image_style
+                                )
                             else:
                                 img_url = generate_image_replicate(scene_summary, st.session_state.protagonist)
                                 
@@ -503,40 +586,17 @@ def main():
                     
                     # Display image
                     st.image(st.session_state[f"image_url_{i}"], use_column_width=True)
-                    st.caption(f"Illustration generated for Scene {i+1}")
-                
-                with col2:
-                    st.markdown(f"<div class='scene-container'><div class='story-text'>{scene}</div></div>", unsafe_allow_html=True)
-        
-        # Full story text
-        with st.expander("View complete story"):
-            st.markdown(f"<div class='story-text'>{st.session_state.story}</div>", unsafe_allow_html=True)
-        
-        # Optional TTS toggle
+                    st.caption(f"Illustration for Scene {i+1} in {st.session_state.image_style} style")
+    
+    # Story text display that changes based on selected tab
+    if "story" in st.session_state:
         st.markdown("---")
-        st.subheader("🔊 Narration")
+        st.subheader("📚 Your Personalized Story")
         
-        if st.checkbox("Include audio narration"):
-            # Check if TTS is imported
-            try:
-                from google.cloud import texttospeech
-                with st.spinner("Generando audio..."):
-                    if "audio_file" not in st.session_state:
-                        audio_file = generate_audio_tts(st.session_state.story)
-                        st.session_state.audio_file = audio_file
-                    
-                    if st.session_state.audio_file:
-                        st.audio(st.session_state.audio_file)
-                    else:
-                        st.warning("Could not generate audio. Please check your Google Cloud configuration.")
-            except ImportError:
-                st.warning("""
-                The narration feature requires Google Cloud Text-to-Speech.
-                
-                To enable this feature:
-                1. Install the library: `pip install google-cloud-texttospeech`
-                2. Configure your Google Cloud credentials
-                """)
+        # Only display the currently selected scene
+        selected_index = st.session_state.selected_tab
+        st.markdown(f"### Scene {selected_index + 1}")
+        st.markdown(f"<div class='scene-container'><div class='story-text'>{st.session_state.scenes[selected_index]}</div></div>", unsafe_allow_html=True)
         
         # Download options
         st.markdown("---")
@@ -554,6 +614,10 @@ def main():
         # This is just a placeholder - in a real app you'd need to implement image downloading
         with col2:
             st.info("Image downloading will be available in a future version.")
+        
+        # Full story text
+        with st.expander("View complete story"):
+            st.markdown(f"<div class='story-text'>{st.session_state.story}</div>", unsafe_allow_html=True)
 
 # Run the app
 if __name__ == "__main__":