from PIL import Image
import requests
import gradio as gr


from transformers import BlipProcessor, BlipForConditionalGeneration

model_id = "Salesforce/blip-image-captioning-base"

model = BlipForConditionalGeneration.from_pretrained(model_id)
processor = BlipProcessor.from_pretrained(model_id)

def launch(input):
    inputs = []
    for file in input:
        inputs.append(processor(Image.open(file), return_tensors="pt"))
        
    out = model.generate(**inputs)    
    return processor.decode(out[0], skip_special_tokens=True)

description = "Simple BLIP test app for image captioning."

iface = gr.Interface(launch, description=description, inputs=gr.inputs.File(file_count="multiple"), outputs="text")
iface.launch()