OpenAI's products and DALL-E 3 can't really transform existing images. GPT-V lets it see images, and DALL-E 3 can make images. But they aren't directly connected.
> We'll analyse your photo, describe it in words, and generate a new image from scratch.
I've tried the same technique to get around "copyright policy" refusals, but never get anything close to what I want.
Those that are complaining about the lack of resemblance, may have missed the text explaining how this works: "We'll analyse your photo, describe it in words, and generate a new image from scratch."
Mine was also way off, but if I think about how an AI would describe the photo and then regenerate it, then it's probably pretty close to that.
Mine kinda sucked. First, it made an avatar with two almost identical heads in it. I have a high forehead, the two people in the avatar had full, thick hair. I have brown eyes, the avatar twins have large, deep blue eyes. The only thing in common with my image was that the avatar people have a salt and pepper beard, like I do.
They allude to it but do not fully explain. I am assuming they are also using the vision model to describe the input photo. I would not have initially thought that you given the text description of an image you could recreate it well enough for fun avatars...it makes sense but I did not initially think about it.
Your pricing is absolutely ridiculous. I ran a full image generation site (like Leonardo), so I know what it costs to run models/generate images. I'm shocked people are willing to pay you that much money. Your margins must be like 99%
I have tried it and I have expected that it would look similar to the selfie, but the output does not look like me. It looks more like a random stock photo. Perhaps that was the intention?
I have to try three times because it says something went wrong. Then it generates an image that looks nothing like the one I uploaded. And then the next time I try it says I made too many requests. So, it would be cool if it worked I guess.
