Update app.py

app.py

@@ -4,6 +4,8 @@ from transformers import BlipProcessor, BlipForConditionalGeneration
 from PIL import Image
 import easyocr
 
+from transformers import pipeline
+
 
 
 # Get Hugging Face Token from environment variable
@@ -15,6 +17,9 @@ login(token=hf_token)
 # Load the processor and model
 processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-large")
 model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-large")
+
+
+pipe2= pipeline("image-to-text", model="nlpconnect/vit-gpt2-image-captioning")
 import gradio as gr
 from diffusers import DiffusionPipeline
 import torch
@@ -24,6 +29,7 @@ import spaces  # Hugging Face Spaces module
 pipe = DiffusionPipeline.from_pretrained("stabilityai/stable-diffusion-3.5-medium")
 device = "cuda" if torch.cuda.is_available() else "cpu"
 pipe.to(device)
+pipe2.to(device)
 model.to(device)
 
 
@@ -55,7 +61,8 @@ def generate_caption_and_image(image):
     inputs = {key: val.to(device) for key, val in inputs.items()}
     out = model.generate(**inputs)
     caption = processor.decode(out[0], skip_special_tokens=True)
-    prompt = f'''Create a highly realistic clothing item based on the following description: The design should reflect {caption}, featuring a highly realistic and modern piece of clothing that incorporates stylish and high-quality textures, exuding sophistication with realistic fabric lighting and fine details, subtly hinting at {selected_fabric}, with a {selected_pattern} motif and a {selected_textile_design} style.'''
+    caption2 =pipe2(img)
+    prompt = '''f'''Create a highly realistic clothing item based on the following descriptions: The design should reflect {caption1} and {caption2}, blending both themes into a single, stylish, and modern piece of clothing. Incorporate highly realistic and high-quality textures that exude sophistication, with realistic fabric lighting and fine details. Subtly hint at {selected_fabric}, featuring a {selected_pattern} motif and a {selected_textile_design} style that harmoniously balances the essence of both captions.''''''