Currently, we don't offer control over the size of the images output from models. This is because each model imposes its own restrictions on image sizes. These restrictions are largely due to the current state-of-the-art algorithms that process images using ML.
If you are interested in generating larger images, you have two options
- Use the ESRGAN image-upscaling model to synthetically upscale the output of another model x4.
- Wait for image generation models to improve. This may sound like a silly answer, but the speed at which these algorithms are being iteratively improved is remarkable. It's not uncommon for an algorithm to be improved by the research community over the course of several months. We're adding new models as fast as we can, and soon these models should be able to output higher resolution images. In the meantime you can learn how to port your own models here.