from PIL import Image
from transformers import BlipProcessor, BlipForConditionalGeneration
import gradio as gr

# Initialization of the BLIP processor and model
processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-large")
model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-large")

def generate_captions(image, text=""):
    # Convert the uploaded image to PIL Image
    raw_image = Image.fromarray(image).convert('RGB')
    
    if text:  # Conditional image captioning
        inputs = processor(raw_image, text, return_tensors="pt")
    else:  # Unconditional image captioning
        inputs = processor(raw_image, return_tensors="pt")
    
    # Generate captions for the image
    out = model.generate(**inputs)
    caption = processor.decode(out[0], skip_special_tokens=True)
    
    return caption

# Gradio Interface
iface = gr.Interface(
    fn=generate_captions,
    inputs=[
        gr.Image(label="Upload/Drag Image"),  # Removed the 'tool' argument
        gr.Textbox(label="Conditional Text (optional)", placeholder="Enter conditional text (optional)...")
    ],
    outputs=gr.Textbox(label="Generated Caption"),
    title="BLIP Image Caption Generator",
    description="This app generates captions for uploaded images. You can also provide conditional text to guide the caption generation."
)

if __name__ == "__main__":
    iface.launch()