P: Outpainting / Extend Image

[brianc83673708]

not sure of some of my wording, but using Comfyui for Stable Diffusion, I've been generating wider-angle shots to allow for custom cropping.  To do this, I basically outpaint and merge after the 14th pass then do 6-10 more passes of that image to create unity.  this is all automated.  It would be a simple implementation that could have a huge impact. 

[Tracy.H]

Have you checked out this video?

[brianc83673708]

I have, and that basically shows you have all the pieces ready, but my point is less of an "editing later" suggestion.  the way the workflow functions is as a high quality outpainting tool during the actual image generation that creates an image with more continuity than is available with pure outpainting.  

keep in mind that I don't know how the "steps" work in this particular Ai, so I'll talk in Stable Diffusion terms.  In this case, an image would go though 20 diffusion steps to become whole:

Basically the workflow (though I have automated)-

1- run 14 steps of the prompt

2- automatically outpaint the prompt by 300 pixels per side

4- Run the final 6 steps using the outpainted image as if it's Image-to-Image.

5- Batch step 4

I also sometimes do this with just a content aware crop in photoshop then running image to image in Automatic1111 but either way, it creates a final product that has a ton of usability for composition.

[Dimotion]

So I believe I have seen this feature in other AI programs, but basicly when it generated a couple of pictures you can choose to extend one of those photos to make it bigger, like to add more to the picture it originally made. Seems like a cool idea to me because when I see an AI generated image, it sometimes has a little detail in the corner or elsewhere around the edges and in my mind I want to see what that might be, so with this feature you are able to do so. And maybe you can even give it a new prompt for the new version of the first one! Let me know what you guys think of this idea!

[Tracy.H]

It sounds like you are interested in the extend image feature. Have you checked out this video?

[jessica.kantak1]

I design paid ads for an ad agency. I currently have to resize my 1:1 designs into other sizes such as 9:16 and 2:3. It would be great to have a feature in Photoshop or something that takes my files and the layers and resizes them appropriately. 

Merging your thread here so we can track the demand for certain features.

[Ghanbarifard]

it must be very easy to generate and add space to all sides. maybe a crop tool or expand tool

Merging your thread here so we can track the demand for certain features. You can upvote the original post to let us know you want this feature.

[Wyatt262088264m8m]

I think it would be very useful and convenient to be able to change the aspect ratio of an image that you have already generated and like by adding to the background, inside of Firefly instead of having to use another app, or generate a new image with the desired ratio. I have several favorited images that I would like in a different aspect ratio, and I know that the AI would have no trouble adding on a little to the outside, but there is no clear way to do this if one exists at all.

Inside of generative fill there is a background button, and I thought it could add more background allowing me to change the aspect ratio without distortion or cropping, but it only removes background. I think these feature would be an execellent to both the generative fill and as an option on the text to image without the text to image coming up with a completely new image. 

Merging your thread here so we can track the demand for certain features. You can upvote the original post to let us know you want this feature.

[María Mercede30614739n36q]

[BK Jaffna]

Need to extend the image boundries based on the nearest colour for a prefered particular shape, for example like attached God as point of light, need the orange colur to be extedn to make it as a rectangular image, thanks a lot for the AI and thanks in advance.

Om Shanthi