I think there is a lot of money to be made in building "human accessible" frontends to AI tools, so I'm not surprised that his ChatGPT wrapper product does well.
Setting up Midjourney, for example, is such a convoluted mess, where you have to create an account on another service (Discord) before creating a Midjourney account, then understand the different chat channels and figure out where the best place to use the app is (private DM with Midjourney bot.) And that's all before even generating an image! Even their website looks like someone's art project, not a cutting-edge image generation service: https://www.midjourney.com/
Leonardo.ai is much better in this regard. I hated midjourney, and if they don’t change this ridiculous way of interacting with the app they will fail.
Meh, I still am not convinced that the blank chat box is a good UI. You still need to know what to say and how to say it. Too much implicit knowledge is required.
You should watch the dall-e 3 trailers. The whole leap forward is that you can just write a description of the scene, the more descriptive the better. The coherence and accuracy seems to be far better than MJ or even SDXL. There's less prompt fu nonsense (aka 4k, hdr, trending on art station, etc) than ever.
Setting up Midjourney, for example, is such a convoluted mess, where you have to create an account on another service (Discord) before creating a Midjourney account, then understand the different chat channels and figure out where the best place to use the app is (private DM with Midjourney bot.) And that's all before even generating an image! Even their website looks like someone's art project, not a cutting-edge image generation service: https://www.midjourney.com/