Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Ask HN: Any tool to generate AI images with exact same uploaded product images?

news.ycombinator.com

Post reply on HN
Print view View original

11–20 of 26 posts

Re: Ask HN: Any tool to generate AI images with exact same uploaded product images?

#11

Post by codetrotter »

mg wrote: post #10
sodiumtech wrote: post #9

Earlier quoted context omitted.

sure. For example check this cap image https://pngtree.com/freepng/deep-blue-peaked-cap_6103980.htm... Now If i want to generate an image of a boy wearing this same cap how can it be done.

In that image, a little bit of the backside of the cap is visible. The dark border at the right bottom. So as I understand it, the AI would have to figure that out on its own and remove it before it adds the boy to the image? Also, that image has watermarks all over it. Does that mean the AI has to detect and remove those? The perspective of the cap is rather unusual for a photo on a human. As it would cover the eyes…

> Also, that image has watermarks all over it. Does that mean the AI has to detect and remove those?

I don’t think that’s what they meant. I think they mean they will use photos of products that they themselves (or the factories they buy it from) provide, without watermarks, and they want to add generated things like people etc into the photo while keeping the product itself the exact same as it was

Re: Ask HN: Any tool to generate AI images with exact same uploaded product images?

#13

Post by throwup238 »

sodiumtech wrote: post #3
muzani wrote: post #2

Would Dream Studio do what you want? https://beta.dreamstudio.ai

Thanks but it doesnt work. It takes the reference from image of the product and creates similar images. What I am looking for is to use the same product image

If you're willing to put in a little work it's possible, but you're going to have to get your product rendered or photographed at the angle you want it to appear in the photo, then mask the product while having the AI generate the rest of the image.

I've done this with some success using t-shirt mockup templates to get the color, shadows, folds and creases in the clothing right then regenerated everything around the shirt.

Re: Ask HN: Any tool to generate AI images with exact same uploaded product images?

#18

Post by lucb1e »

> lets say we have a cap and I want to generate a human image wearing that same cap

Do you have an image of the cap in the right orientation, as it would also appear as it is sitting on someone's head?

If not, any algorithm is necessarily going to have to invent what the cap looks like from another angle, making up details on any previously-hidden side and guessing at the depth of different parts of the still image in order to rotate it into the right orientation

If yes, crop it out and paste it onto the target head

Re: Ask HN: Any tool to generate AI images with exact same uploaded product images?

#19

Post by swatcoder »

Why do you specifically want to use AI or even care how its done?

If you have to do this for only a few things, you can do it for a pittance through a service like Fiverr.

If you have higher or recurring volume, you can employ an offshore company to do this for a very modest monthly retainer.

Just let the provider on the other end figure out the tool that makes it cost effective for them -- AI or otherwise.

Re: Ask HN: Any tool to generate AI images with exact same uploaded product images?

#20

Post by LarsDu88 »

Stable Diffusion Dreambooth.

The basic idea is you find a rare unused token (like say f$#sdafad) and then finetune your image generation model with a specific set of images (say 20 images of your red cap at various angles) while telling it that f$#sdafad is the same thing as your red cap.

Then you can start prompting "f$#sdafad resting on the head of a monkey" and your cap will appear on a monkey's head.

The problem with this technique is the finetuning part. Finetuning can take minutes to hours depending on how many gpus you have and needs to be done individually for every new "token" you want to map to a specific individual or object you want to add to your pre-trained model.

Another strategy is to use some kind of autocropping strategy + generative infill. You can take a semantic segmentation model like Meta's "Segment Anything", then use it to segment out the item of interest manually (perhaps a UI could be built to make this a one-step process). Then take the mask and do a generative infill using some sort of image generation model like stable diffusion.

Post reply on HN