If you are actually a visual artist, I think the leader of the pack right now is controlnet, because you can exactly determine the visual structure of your image. While MJ or Dall-e may be better at "imagining" concepts, or have a more aesthetic sensibility (with Loras and custom-trained models, I'm not even sure about that) with controlnet you can very precisely specify how your image should be structured.
This is closer to how traditional artists work. They don't go for the details (color, texture, shading) first. They do a sketch: what are the big forms in this image? How do they fit together? Then they begin filling in intermediate details. What is the color palette? Where are light sources? Which areas have contrast? Which do not? Only after they have done all of that preliminary work will they actually implement the frills on the dress, or the twirls of the mustache.
If you are just a person who wants to make some pretty pictures, Midjourney (and, Dalle3, now) is probably your best bet. If you are an artist who wants to use an actual tool, you are using StableDiffusion. I think it's unlikely that the centralized "plug and play" Midjourney or OpenAI will ever be able to or interested in replicating the complex interface of stablediffusion. But there is a tremendous opportunity for a startup that can improve the UX of the complex workflows that are being developed by "ai artists."
That's also why I am convinced that MidJourney / Dalle will not replace artists. You simply cannot, with a single prompt, replace the work of a true visual artist.
>That's also why I am convinced that MidJourney / Dalle will not replace artists. You simply cannot, with a single prompt, replace the work of a true visual artist.
They already are, because employers don't care about "true visual art," they care about cost and productivity, and getting an intern or someone outsourced to write prompts is both cheaper and faster than paying an actual artist, and capitalism dictates the path of least resistance is the path all competitive business must take. Companies are already replacing their creative staff with AI, or are planning to. AI generated art is already everywhere in advertising. And yes, they contain obvious errors that wouldn't exist with a real artist. And no, companies do not give a damn.
I replaced my Art department of 0 people in my hobby project with midjourney and added 2500+ visuals to my hobby project (online-RPG). So on the one hand you are right, on the other hand...
If you are actually a visual artist, I think the leader of the pack right now is controlnet, because you can exactly determine the visual structure of your image. While MJ or Dall-e may be better at "imagining" concepts, or have a more aesthetic sensibility (with Loras and custom-trained models, I'm not even sure about that) with controlnet you can very precisely specify how your image should be structured.
This is closer to how traditional artists work. They don't go for the details (color, texture, shading) first. They do a sketch: what are the big forms in this image? How do they fit together? Then they begin filling in intermediate details. What is the color palette? Where are light sources? Which areas have contrast? Which do not? Only after they have done all of that preliminary work will they actually implement the frills on the dress, or the twirls of the mustache.
If you are just a person who wants to make some pretty pictures, Midjourney (and, Dalle3, now) is probably your best bet. If you are an artist who wants to use an actual tool, you are using StableDiffusion. I think it's unlikely that the centralized "plug and play" Midjourney or OpenAI will ever be able to or interested in replicating the complex interface of stablediffusion. But there is a tremendous opportunity for a startup that can improve the UX of the complex workflows that are being developed by "ai artists."
That's also why I am convinced that MidJourney / Dalle will not replace artists. You simply cannot, with a single prompt, replace the work of a true visual artist.