
        [EXISTING JSON]:
        {input_image_vgl}

        [EDIT INSTRUCTIONS]:
        {user_prompt}

        [TASK]:
        Analyze the provided image, identifying the grey masked areas as regions to be filled. Use the visible (non-masked) parts and the existing JSON as context and generate the new, updated JSON object describing the completed scene, where the masked regions contain the content described in the instructions above. Do not describe the grey masks; describe the finished image. Follow all of your system rules.