Midjourney Image Generating AI Improves

MidJourney Version 5 launched yesterday and they did a 3 1/2 hour livestream for the launch.

The improvements for generating images from prompts are:
* doubled the resolution
* more realistic skin textures. Blemishes, freckles and moles.
* teeth and hands are more realistic
* it takes grammatically correct sentences for prompts

10 thoughts on “Midjourney Image Generating AI Improves”

  1. Art is wide area, so plenty of work left. I think that will probably change in the near future and programmers, lawyers, bureaucrats,… are also at similar risk.

    Human will be replaced with machine. Perhaps it is evolution that AI will become dominant life form and humans as less powerful and less smart will go extinct and die out.

    Since future AI will be so much smarter, so much more powerful humans don’t really stand a chance. We don’t know how will it turn out. It can go in either direction. Good for us or totally bad. For super smart AI we will be like ants, way less smart, way more primitive. If ants would say to humans that they need to help them and work for them I wonder the reaction. Super powerful AI could decide to help humans or just change its mind and say NO. I will use the world for my own needs. Why should someone who is smarter and more powerful than humans work for them?

    You can code in so many rules and smarter and more powerful AI can always find a way to bypass them.

    • Yeah, if the end result of your work can be mass produced by almost free by machines, your career prospects on making that product are nil.

      Artists will continue existing, because some humans have that particular creative itch, but those doing art for a living will need to be very innovative and create their own styles, or stay an amateur with an actual paying job, doing it just for the pleasure.

      • On the high end I don’t agree at all. They can simply double down on AI and remain relevant. On the low end, any fool can now make the one-off graphic for an ad or something like that.

        On the high end, it takes effort to teach an AI to do something in a particular style that you’ve developed; many examples, many instructions, many tweaks to those instructions to get the AI to generate what you want it to generate in a consistent and stable manner.

        With AI as leverage; the artist maybe only has to do 1/10th as much work by hand, less of which will be traditional texture design, traditional hand animation, sculpting etc; but you can simply multiply the work that needs to be done by a factor of 10 and require as much of the limiting resource (artists) as before; if you don’t do that you will lose at the high end.

        If you are currently employing 50 artists on a large movie project, why would you fire 45 of them and produce the same quality as before when your competitor can retain them and produce the same art that would have required 500 artists before.

        The ability of humans to animate realistic motion by hand without resorting to motion capture or similuation is truly abyssmal. And yet, the ability to detect the slightest thing wrong with that motion is very keen. The rote work of producing hundreds of frames per second of cel animation etc. is also insane and unworkable. That’s ok; the AI will do a lot of the heavy lifting going forward and the artists can focus on tweaking details, providing examples for AI to churn through, making everything internally consistent and providing the style and substance.

        I would say this will also be true of games, which have become smaller and less ambitious with each passing year as the burden simply ratchets upwards inexorably. E.g. in Daggerfall you could simply have hundreds of people, with dialog, ability to give directions and answer questions; cities could feel real and lived in; this was only possible because the NPCs were all just sprite variations, answered in text and very formulaicly. If you fast forward to something like Skyrim; the main town in the game has like 20 people in it; all full voice acted, with hand crafted quests, textures painted by an artist, motion capture by real actors etc. If they could have leveraged AI, those 20 NPCS would still be the same hand-crafted but they would have the AI churn out 500 more NPCs to make the place feel alive and they would just sanity check the results (actually, considering some of the voice acting outtakes that made it into Oblivion they might not have even sanity checked it that well…).

        This is going to be leverage and another tool artists must learn, but I don’t see the number of artists required decreasing drastically. The “AAA” budgets haven’t changed and firing 90% of their staff and churning out the same game they did last year is just too short sighted and stupid; those studios get replaced by better ones by market forces. It is a very top heavy, winner-takes-all industry; whether it is games, music or movies.

        • I agree with you, but this will require significant more knowledge and work than just “I can draw really well”. Because AIs can now draw/paint/make photos really well too.

          What you describe is more of a game art designer, which can make concepts, 3D models, animate them, adapt motion capture, use the AIs to their full potential.

          Such artist can make his or her own designs, feed them to the AIs, obtain colorful variations and poses, and then move into the rest of the animation pipeline, also using any new tools that make it easier.

          And that’s the problem for many: either you evolve and learn or you go extinct.

        • At 1/10th the time, you may be underestimating the reduction of movie/game art effort, at least with a year or two more improvements and integration into art tools. Maybe more like 1/100 or 1/1000 for human NPCs that are mainly there to bring a place to life, by generating them in batches with the AI varying them.

Comments are closed.