app.py

#imagetext-to-text
import gradio as gr
import base64
from huggingface_hub import InferenceClient
#client = InferenceClient('Qwen/Qwen2.5-VL-7B-Instruct')
#client = InferenceClient("mistralai/Pixtral-12B-Base-2409")
client = InferenceClient('meta-llama/Llama-3.2-11B-Vision-Instruct')

def imageDescription(image, prompt):
 image_path="image.png"
 image.save(image_path)
 with open(image_path, "rb") as f:
   base64_image = base64.b64encode(f.read()).decode("utf-8")
 image_url = f"data:image/png;base64,{base64_image}"
 output = client.chat.completions.create(messages=[
       {
           "role": "user",
           "content": [
               {
                   "type": "image_url",
                   "image_url": {"url": image_url},
               },
               {
                   "type": "text",
                   "text": prompt,
               },
           ],
       },
   ],
 )
 return output.choices[0].message.content


with gr.Blocks(theme=gr.themes.Citrus()) as demo:
   with gr.Row():
     with gr.Column():
       #an image input
       image=gr.Image(type="pil", label="upload an immage")
     with gr.Column():
       prompt = gr.Textbox(label="What would you like to know about this picture?",scale=1)
       describe_btn = gr.Button("Describe the image",scale=1)
       output = gr.Textbox(label="Description",scale=1)
       #sending two inputs to imageDescription function
       describe_btn.click(fn=imageDescription, inputs=[image, prompt], outputs=output)
demo.launch(debug=True)