Dalle-3 has arrived!
Still not a Midjourney killer
I decided to plunk down the $20 for a month of ChatGPT+ so I could test the Dalle-3 integration and so far I’m really impressed! While the image quality is not quite Midjourney level, it is a huge leap forward in how we write prompts. You can read my longer comparison in this medium article. My goal is to try and use it every day for the next month and really test it’s capabilities. I’ll have another write-up next week.
It’s important to mention that the the Dalle-3 in BING is not the same as the Dalle-3 in ChatGPT. I don’t know how they do this, but the images from the latter are much better.
Dalle-3 vs. Midjourney: Detailed Comparison
Wei Mao goes into much greater detail in his head to head comparison of Dalle-3 and Midjourney. He might be a better prompter than me or perhaps he re-rolled a few times, but he is getting amazing images out of Dalle-3 and in most of his comparisons, it beats Midjourney. I tried copying his prompts and my results weren’t as good. Since his opinion is different than mine, it’s worth a look.
Using ChatGPT and Dalle-3 Together to Recreate Images
One major flaw in Dalle-3 is that it doesn’t have an image upload feature. In Midjourney and the New Firefly, you can upload a reference image and with a little tinkering you can get something similar, but not identical to the original. Wei Mao (again) has a fascinating work flow where you first give ChatGPT specific instructions for how you want an image described, you upload an image and it will produce four different prompts. You then copy these prompts, switch over to Dalle-3 and past them to produce sometimes remarkable results. I know it sounds tedious, but a brilliant solution if you’re willing to invest the time and it’s worth a look.
Real Designers Using AI in Their Real Design Jobs
Are you still on the fence about using generative AI tools in your work as a designer?
Are you worried that you might miss this train if you don’t act soon? Why not read this article from Creative Bloq where they profile several design studios and go over the pros and cons of using AI tools in their real workflow. It’s worth a look.
Adobe Has a Plan. We Can trust Adobe, Right?
Creative Bloq (again) has a detailed report about where Adobe stands in it’s continued expansion of AI tools. What we saw at Adobe MAX last week is just the beginning. While I never thought that I’d use the words ‘Adobe’ and ‘Good Citizen’ in the same sentence, it sounds like someone at Adobe is trying to do the right thing. If you’re curious about artists rights, copyright, digital watermarks, it’s worth a look.
Deep Learning for Product Designers
Spoiler Alert: This is a long article with a technical exercise at the end. If you’re a product designer, I really want you to focus on the first half. Maria Kovalevich makes an excellent case that if you’re in the business of building digital products, it’s in your best interests to understand the concepts of ‘Deep Learning’ and ‘Natural language processing’. While you don’t need to go as far as she did, I think it’s worth a look.
MidJourney Now has a 2X and a 4X Upscaler
Maybe I was on autopilot last night, but I didn’t even notice these buttons. I only discovered this feature in my morning reading. The upscaling happens pretty quickly considering the number of pixels being rendered and there appears to be no loss of quality or weird artifacts. The file sizes are gigantic for the 4X and should be suitable for print work. I’ll be playing with it more throughout the week.
That’s it for this week I guess it was a slow week for AI news, or maybe I’m just getting selective. Either way, I’ll keep trying to bring you the most relevant news every week. I hope you’re having a great time out there and your learning journey is going well.