Failed to obtain a variable-resolution image.

[lucky-liuzhihong]

Great work! I noticed that the paper mentions:

"We find that by appending instructions about the resolution to the input text prompt, such as 'length is: width is:', the model can quickly learn to generate the corresponding code according to the specified number of rows and columns."

However, when I add “length is 960: width is 640” to the prompt when using the demo model on Huggingface, the generated image is still 512×512. Did I make something wrong?

Could you please clarify? Thanks!

[wjf5203]

Thank you for your attention! Generating multi-resolution images requires adding prompts specifying width and height during training. The currently released model was trained at a fixed resolution. While we have validated the feasibility of multi-resolution models in early stages, we will consider releasing such weights if there is a need.