But what if the model you're using doesn't have image processing capabilities?