I never used Dall-E, well I tried a few times way back, but it never amounted to anything usable. Seems V3 is a huge improvement! Going to give this a whirl. Appreciate the explanations, David. I always choose "short runtime" (fewer steps) because of credits. Still don't understand the Sampling Method differences, EULER, DPM, LMS, DPMPP_2PP (Default) etc... There are a lot of parameters and negative prompts and stuff where writing a good prompt is the biggest challenge and then being lucky and getting a good seed, at least with Stable Diffusion. Dall-E 3 seems to get rid of all that cryptic pre-work.
I completely agree, the original Dall-E was pretty mundane. I didn’t even use all of the tokens I bought lol. I’d love to see your results with V3, be sure to send links here!
As for the Stable Diffusion parameters, I never spent the time learning how to properly manipulate these. I also found the prompting and ‘getting lucky’ a bit annoying.
I actually was trying to get a futuristic smart home image during a full moon Halloween night for my Protocol H.25 story but opted to use one of my own images of a pumpkin we carved, instead.
Yeah, I got a similar one. The issue I have with all of them is they look too “generated” if you know what I mean, maybe have to add certain artist style modifiers to get a more authentic look... or use my own pictures ;)
I wish I could zoom in to see the ads better! But from what I can see, you’re correct how the quality of fun and helpful changed to godlike in the futuristic ads.
It's definitely tougher to see in the mobile app, but you should still be able to touch the graphic to make it expand, and then be able to zoom in at different regions.
Fully agree; DALLE has quite the opinion of itself in the future.
Hi Nathan, thank you! I had a lot of fun making the art for this one. The tools us small authors now have access to is paradigm shifting. Individuals can now have the full feel of a full publication. And we’re only moonlighting! I’m anxious to give this a shot for some science-based applications next.
What tools is your brother using for giving the appearance of motion?
I like how DALL-E made the mammoth essentially dance while also changing the background of the image, which suggests it can't recreate an image from a previous reference which is kind of strange. You would think it could at least copy itself.
I never used Dall-E, well I tried a few times way back, but it never amounted to anything usable. Seems V3 is a huge improvement! Going to give this a whirl. Appreciate the explanations, David. I always choose "short runtime" (fewer steps) because of credits. Still don't understand the Sampling Method differences, EULER, DPM, LMS, DPMPP_2PP (Default) etc... There are a lot of parameters and negative prompts and stuff where writing a good prompt is the biggest challenge and then being lucky and getting a good seed, at least with Stable Diffusion. Dall-E 3 seems to get rid of all that cryptic pre-work.
I completely agree, the original Dall-E was pretty mundane. I didn’t even use all of the tokens I bought lol. I’d love to see your results with V3, be sure to send links here!
As for the Stable Diffusion parameters, I never spent the time learning how to properly manipulate these. I also found the prompting and ‘getting lucky’ a bit annoying.
I actually was trying to get a futuristic smart home image during a full moon Halloween night for my Protocol H.25 story but opted to use one of my own images of a pumpkin we carved, instead.
Have only tried real quick just now: https://www.bing.com/images/create/lone-2-storey-house2c-futuristic-style2c-sustainable/652a564187704a7a94e955c1718fb35f?id=5MOW8c%2ba2YaLXIPKyM0eHQ%3d%3d&view=detailv2&idpp=genimg&FORM=GCRIDP&mode=overlay
"no lights inside house" is not something it understands, will need to rewrite that prompt.
I hadn't tried the Bing image generator yet! But it's pretty clear many of these do not understand context or specific details.
I just tried your prompt with DALLE 3 and the image looks pretty darn good. I'm attaching the photo via link : https://ibb.co/5hX2DTN
I still think your photography beats any of these (;
Yeah, I got a similar one. The issue I have with all of them is they look too “generated” if you know what I mean, maybe have to add certain artist style modifiers to get a more authentic look... or use my own pictures ;)
I agree, there is still something a bit uncanny about AI generated art and text. It's great to have a hobby that matches so well with your work.
Thanks, David. Only need the work/hobby to be sustainable now 😅
I wish I could zoom in to see the ads better! But from what I can see, you’re correct how the quality of fun and helpful changed to godlike in the futuristic ads.
It's definitely tougher to see in the mobile app, but you should still be able to touch the graphic to make it expand, and then be able to zoom in at different regions.
Fully agree; DALLE has quite the opinion of itself in the future.
Super impressive, David! Great to see the exploration of ideas here.
Personally, I'm ok with those 2070s/2080s posters ;)
My brother has been using Stable Diffusion a lot. I also see they have some quite impressive animations now (well, relatively).
Hi Nathan, thank you! I had a lot of fun making the art for this one. The tools us small authors now have access to is paradigm shifting. Individuals can now have the full feel of a full publication. And we’re only moonlighting! I’m anxious to give this a shot for some science-based applications next.
What tools is your brother using for giving the appearance of motion?
I like how DALL-E made the mammoth essentially dance while also changing the background of the image, which suggests it can't recreate an image from a previous reference which is kind of strange. You would think it could at least copy itself.