Update app.py

app.py

@@ -1,17 +1,13 @@
 # Import libraries
-
 import gradio as gr
-from transformers import pipeline
+from transformers import GPT2LMHeadModel, GPT2Tokenizer
 from gtts import gTTS
 from io import BytesIO
 from PIL import Image
-from diffusers import DiffusionPipeline
-
-# Use a pipeline as a high-level helper for text generation
-vision_alpha_pipe = pipeline("text-generation", model="NousResearch/Nous-Hermes-2-Vision-Alpha")
 
-# Initialize DiffusionPipeline for text-to-image
-image_generation_pipe = DiffusionPipeline.from_pretrained("stabilityai/stable-video-diffusion-img2vid-xt")
+# Load GPT-2 model and tokenizer
+model = GPT2LMHeadModel.from_pretrained("gpt2")
+tokenizer = GPT2Tokenizer.from_pretrained("gpt2")
 
 # Set the context for the storyteller
 messages = [{"role": "system", "content": "You are a magical storyteller, creating wonderful tales for kids. Make them imaginative and full of joy!"}]
@@ -20,39 +16,45 @@ messages = [{"role": "system", "content": "You are a magical storyteller, creati
 current_page = 0
 
 # Define the Storyteller function
-def StorytellerNous(character, child_name, lesson_choice, tell_story, _):
+def StorytellerGPT(character, child_name, lesson_choice, tell_story, _):
     global current_page
-    
+
     # Set the characters and lesson based on user choices
     character_info = f"Once upon a time, {child_name} met {character}. "
     lesson_info = f"Today's lesson is about {lesson_choice}. "
-    
+
     messages.append({"role": "user", "content": tell_story})
-    
-    # Generate story using Nous-Hermes-2-Vision-Alpha
+
+    # Generate story using Hugging Face's GPT-2
     input_text = character_info + lesson_info + tell_story
-    story_reply = vision_alpha_pipe(input_text, max_length=150, num_return_sequences=1, no_repeat_ngram_size=2, top_k=50, top_p=0.95)[0]['generated_text']
+    input_ids = tokenizer.encode(input_text, return_tensors="pt")
+    story_reply = model.generate(input_ids, max_length=150, num_return_sequences=1, no_repeat_ngram_size=2, top_k=50, top_p=0.95)[0]
+
+    # Decode the generated sequence
+    story_reply = tokenizer.decode(story_reply, skip_special_tokens=True)
 
     messages.append({"role": "assistant", "content": story_reply})
-    
+
     # Convert text to speech
     tts = gTTS(text=story_reply, lang='en', slow=False)
     audio_io = BytesIO()
     tts.save(audio_io)
     audio_io.seek(0)
 
-    # Convert text to image using DiffusionPipeline
-    image_reply = image_generation_pipe(story_reply)
-    
+    # Convert text to image
+    image = Image.new("RGB", (300, 300), (255, 255, 255))
+    image_path = "/path/to/output/image.png"
+    image.save(image_path)
+
     # Display the story on separate pages
     story_pages = story_reply.split("\n\n")  # Split the story into pages
     current_page = min(current_page, len(story_pages) - 1)  # Ensure the current_page is within bounds
-    
-    return story_pages[current_page], Audio(data=audio_io.read(), autoplay=True), image_reply
+
+    return story_pages[current_page], Audio(data=audio_io.read(), autoplay=True), image
 
 # Create the Gradio Interface with styling
 demo = gr.Interface(
-    fn=StorytellerNous,
+    fn=StorytellerGPT,
     inputs=[
         gr.Textbox("text", label="Child's Name"),
         gr.Dropdown(["unicorn", "dragon", "wizard"], label="Choose a Character"),