Image Editing Capabilities

[bcaffee]

Hello, thank you for publishing this work! I was wondering if you can provide code for image editing, i.e. input image and text and output image.

Edit: I see that it is mentioned here that the model has only been trained for text-to-image generation but is there any way we can reproduce the example above?

[0001Henry]

the same question :)

[ZKCCZ]

the same question :)

[EladWarshawsky]

same lol

[bcaffee]

Well, it's been a month. My theory is it'll be at least a few more months before we get a response. 😆

[chaochao0]

the same question :)

[onepeachbiubiubiu]

Well, it's been a month. My theory is it'll be at least a few more months before we get a response. 😆

I guess no, the paper does not show image editing cases.

[wjf5203]

Thank you for your attention! We have successfully tried multi-round generation on Liquid, but the very long context affects the training efficiency, so the model currently released is trained on a single image for generation and understanding, which our main evaluations focus on.
We have seen that the community is increasingly interested in multi-round dialogue generation and image editing. We are working on this to add multi-round data into pretrain and will release it as soon as possible!

[EladWarshawsky]

Thank you for your attention! We have successfully tried multi-round generation on Liquid, but the very long context affects the training efficiency, so the model currently released is trained on a single image for generation and understanding, which our main evaluations focus on. We have seen that the community is increasingly interested in multi-round dialogue generation and image editing. We are working on this to add multi-round data into pretrain and will release it as soon as possible!

Thank you for the update

[ChunChenLin]

the same question :)