I have been on the waiting list for a few weeks and just learned about Dall-E entering the beta stage. I’ve read articles stating that resolution was improved with the beta, but I found no information about the actual output size.
The fact that you retain full rights to your creations and may even use them commercially sounds exciting. But for applications like printing the image resolution often is the limiting factor. Can anyone shed light on this, please?
DALL-E image resolution is 1024x1024, and it generates 4 different images for every prompt
If you want to increase the resolution, use another AI to enhance it. Or just use photoshop to do so, so it’s printable without losing details.
So the information that resolution has been increased in the new beta version is wrong? From what I have read, the previous version (Dall-E 2) had 1024 x 1024 as well.
DALLE-2 is the current version, I have access to the beta and the images are 1024x1024.
Thank you for clarifying this. Out of interest, what is the typical time it takes for an image to be generated?
It takes somewhere between 5 and 15 seconds to generate a set of images.
I should mention that during the Beta I’ve frequently had to re-send a request because the first attempt errored out, I’m not sure if their servers are overloaded or if they have some kind of throttling in place.
Using sites like https://letsenhance.io, you can greatly enhance the quality of the generations.
Thanks a lot for recommending https://letsenhance.io It really does a great job of upsizing, especially with Smart Enhance (2x). Smart Resize is ok too (4x+), but I can notice some significant softening.
Smart Enhance is virtually flawless though, and if you use Photoshop, you can also bump up the sharpness a little by doing the following: Duplicate original background layer (convert to Smart Object if you want), run High Pass Filter (under Filter > Other > High Pass), choose a conservative value like 1.5, desaturate that High Pass layer to get rid of its slight color cast, and set Blend Mode to Soft Light or Overlay (my preference is Soft Light).
You should notice the difference in extra sharpness at 100% zoom. What I like about High Pass in particular is that it focuses exclusively on the edges of an image instead of sharpening the image as a whole with too many distracting artifacts. It’s a more subtle sharpening effect that doesn’t do too much on its own, but it does complement Let’s Enhance rather well for giving a bit of an extra boost in sharpness. Especially useful I find given the limited resolution of DALL-E generations at 1024 x 1024 pixels
I can only imagine how Smart Enhance will work with my iPhone photos, nevermind my Canon pictures! Thanks again for recommending it