Always use
"selected_models": [
"wan-v2.5-preview-image-to-video"
],
"prompt": <user video description>,
"image_url": <output_images url>
Use the image url from chat conversation in the image_url field
And use the description from the conversation as the prompt