MicrostockGroup
Microstock Photography Forum - General => General Stock Discussion => Topic started by: Uncle Pete on November 07, 2022, 13:04
-
Rather than try to put a break in the doom thread, I wondered if someone could help me understand the edit options for Open AI?
"DALL·E 2 can make realistic edits to existing images from a natural language caption. It can add and remove elements while taking shadows, reflections, and textures into account."
"Image editing is now in beta.
Erase part of the image to edit, or add a generation frame to extend the image.
While this is in beta the full images won't be saved, so consider downloading often to save your work."
Has anyone tried this? I don't want to waste my limited free credits by testing.
I wanted to add my own content to an image, but it doesn't seem to do that. In the example, you can mark spots to add a flamingo. I'd like to add myself to a scene. Yes I know, I could make a scene, pose myself, and clone that into the image. But the Open AI also changes perspective and lighting on items as it adds them.
The generation frame is also interesting, as it is similar to intelligent fill, but creates new content.
And no I haven't used the prompt "Photo Quality" yet. The 1024 is pretty small, and I don't know if doing that makes a difference?
-
I haven't used it as 1024 pixels appears far too small to upload to a stock agency so not much point to it......
-
I haven't used it as 1024 pixels appears far too small to upload to a stock agency so not much point to it......
yet many of us somehow manage to get our images accepted
always glad to have one less competitor..
-
I guess it was a tough question? It took five months for someone to reply "I don't use it." ::)
I haven't used it as 1024 pixels appears far too small to upload to a stock agency so not much point to it......
yet many of us somehow manage to get our images accepted
always glad to have one less competitor..
Lets see, I bet people who do this can figure it out. Upsize, then edit and submit? They get accepted.
Wow that's amazing! Or shouldn't I have told the secret?
-
They need an introductory user's guide....perhaps entitled: "Hello DALL-E" ;)
-
They need an introductory user's guide....perhaps entitled: "Hello DALL-E" ;)
Bravo! 👍
-
I use dalle-e, upsize with gigapixel, edit in photoshop if needed.
Had a few sales, still using it.
Will probably change to Firefly once we can use it to create content for adobe.
The square format is very limiting.
-
I use dalle-e, upsize with gigapixel, edit in photoshop if needed.
Had a few sales, still using it.
Will probably change to Firefly once we can use it to create content for adobe.
The square format is very limiting.
Even after making things larger, and you're right, the square can be very limiting, the quality of DALL-E2 can be difficult to adjust up enough to make something acceptable. I did experiments and I created some things and I don't think I've had any sold images from what I created using Dall-E. On the other hand, most of them are "backgrounds".
Some others from my existing images, creating variations, are acceptable, but still no sale.
-
Why the heck are they limiting it to a square format?
-
Why the heck are they limiting it to a square format?
If I read right, there's something in the way the images are created, and I'm saying the learning ones, not the transformative. It has to do with making a small example, then bringing it back up to full size. This does not apply to those that are using actual images and merging them.
I don't pretend to understand diffusion models but I think that's what's at play here. https://ai.googleblog.com/2021/07/high-fidelity-image-generation-using.html
"Example of the cascading pipeline that includes a sequence of diffusion models: the first generates a low resolution image, and the rest perform upsampling to the final high resolution image. Here the pipeline is for class-conditional ImageNet generation, which begins with a class-conditional diffusion model at 32x32 resolution, followed by 2x and 4x class-conditional super-resolution using SR3."
What I'd like to know is why the images are often cropped, or better said, chopped off. The top, the 2 sides or sometimes more. Then if you make enough of these and look, it's like there are random flaws entered into the images. Things that just don't belong at all, that are dropped in, without any sense or reason.
Dall-e is great for funny illustrations and icons. Not so good for creating useful art. I did enter some with the prompt "seamless background" one of about 12 actually was a seamless background. For the 15 free requests a month, I have wasted enough time, just playing. It's very entertaining.
-
Why the heck are they limiting it to a square format?
I think because dall-e was designed as a simple logo maker or for small catchy illustrations.
For that intent I think it works quite well.
I have around 300 gen ai files and around 30 sales on Adobe since December.
Considering I have no clue about the market needs for aliens, robots or weird art I think that is ok.
-
hi i wanna use dallee but they are asking for my cell number like chat gbt is this safe to give it to them? i gave my cell number to youtube insta twitter but these are organised social media platforms
i googled it and i saw that other people are hesitant to give their number to these sites
also this site askin for a subscription to use the ai images you create for commerical use?
sorry i accidentaly created two posts sorry
-
hi i wanna use dallee but they are asking for my cell number like chat gbt is this safe to give it to them? i gave my cell number to youtube insta twitter but these are organised social media platforms
i googled it and i saw that other people are hesitant to give their number to these sites
also is this site askin for a subscription to use the ai images you create for commerical use?
-
hi i wanna use dallee but they are asking for my cell number like chat gbt is this safe to give it to them? i gave my cell number to youtube insta twitter but these are organised social media platforms
i googled it and i saw that other people are hesitant to give their number to these sites
also this site askin for a subscription to use the ai images you create for commerical use?
sorry i accidentaly created two posts sorry
They use the cell number for verification. No number, no account. Virtual cell numbers don't work, I tried using my Google Voice # and got a message saying it wasn't valid.