I'm starting a furniture project with Midjourney. Generating a bunch of ideas in the styles I'm interested in, then choosing the most feasible outputs to bring to life with wood.

As midjourney created output increases over the years, perhaps mj will start to consume its own output as training material. All art will then converge into a Bob Ross style painting of the number 42...

1. One big use I foresee is storyboards (for movies, games) where a director/writer etc creates a series of images (selected by him/her) to quickly and visually convey the idea for a studio etc. Many movies and games re-use known, stereotypical settings, so having that "Artstation" look is not an issue 2. There's no chance of human graphic designers going out of business entirely, because no matter how good an AI image is, the client will ALWAYS want some change ("can you move this hill a bit to the right?"), just like in the past. This won't change. "Tweaking" AI images will exist for a long time.

people blow this whole thing way out of proportion. artists aren't really going anywhere. The tools might change but the artist remains. my advise to everyone is adapt to the new tools and stop being scared about it. If for whatever reason you're scared regular people with no art education now have easy access to creating pictures, go have a look at the pictures that are being created with dall-e 2. like 99% are just random meme images. You're gonna be fine.

I'm amazed that anyone thinks of this in a positive way. Why would you hire an artist in the future if the AI results are good enough? If you need a shoe ad for a website, why hire an art director and artists and wait weeks for one ad that may or may not get attention. Why the effort when you can get an accurate psychological profile of the user and generate your product mixed with what interests him (remember the Cabridge Analytica scandal? ;)). If he's having a good day, it will be funny, if he's having a bad day, it will be uplifting. If he likes dogs, something fun with dogs. If his engagement with ads about dogs drops then cats, athletes, etc will be added. If he likes to experiment then the ads will be very different, if he is aesthetically conservative then they will be very similar. Experimentation costs nothing and happens in minutes. It's an AB test with results in seconds. This way you generate an endless stream of ads made specifically for that person mixed with your product. AI already creates interesting short stories (search the web, they can be great!), images, starts designing games, movies, modeling and animating. The road is still far, but the direction is one. Artists are now helping to train this further by using this tool by choosing aesthetically better variations. At some point only a fraction of people will see the difference between an artist and a machine. Why bother with artists then?

Loved the talk! Thanks for wrapping this up so nicely.

There are some spaces that remain untouched by this: portraits of celebrities promoting products. The film/advertising poster type of work is pretty safe for a while. But album covers/packaging all the artsy type of design will be hit the hardest. Concept art, if a studio had 20 artists generating ideas it will let 10 go and keep the best ones to paint over and use AI to generate more ideas in in a quarter of the time. And one subject that is still to debate is animation. The diffusion method and the space that AI reads information from are not really animation optimal as I understand it. It will probably be able to make trippy loops and transform from one prompt to the other but not really spin the wheels of a car when it moves. (yet ) I am really looking forward to an invitation to dall-e or midjourney hope to get one soon.

Great discussion, one thing I'd like to point out is that you mentioned "this is not a prompt I would've given to describe the outcome". The prompt alone can't generate the outcome in almost any of the examples you've shown. It's through several hundred iterations of iterations of iterations from the original prompt that the final image comes out. For example, yesterday I made some Evangelion themed bank notes with midjourney. The first iteration is mostly just the character painted similarly to how printed money looks like. Only after 200-300 variations did I get a cohesive bank note with the character on it.

"When pixels were still made of wood" I love that 😂😂😂

Funny, i was reading "So Bright the Vision" by Clifford D. Simak, a sci fi writer from the 50s, just after watching your talk, without knowing beforehand what it was about, and a part of this short novel is about writers using "storytellers", machines which create the stories, and write books, the writers are just technicians here to assist the machine, and enter the prompt by choosing the key elements they want in their story. Manually writing is considered ridiculous and almost shameful.

It's really addictive to play with these tools! When you look at Pytorch3D and Differentiable rendering, it doesn't seem too far-fetched to be able to extract the 3D scenes from the 2D images created by these AI models. Once we have the 3D scenes we can make games, movies, and whatnot. It's an exciting time to be in this industry.

Google and Facebook are the biggest advertisement delivery networks out there. Every ad published goes through their pipeline. Both companies now have a tremendous training set. Pair this with the data driven nature of the modern online ads world (click-through rates, conversion, retention), which both companies are masters at and an AI image creator, we will see the downfall of the whole advertisement industry as we know it right now in the five to ten years. These companies will offer any client an all-in-one package from creative to delivery, they will create on-demand-ads catered to every specific user out there in real-time. The recipient won't see the difference. I think it's great that you guys still value concept, taste and craftsmanship, but from my experience in tech and mobile, I would say, that this isn't relevant to the client. He/she cares about KPI and metrics. That's why we see the quality of ads diminishing at a rapid rate lately. Ads are more and more catered to their performance than the ego of a creative mind. I've never been a fan of the concept of originality, I think it's way over-used. I'm really on the side that #everythingisaremix. The trick is/was to pull off unknown references, sample sources, no one else though off to stay fresh. So perhaps the difference between a human and the AI right now, is that a human can enlarge its "training set" (inspiration) on purpose and with a goal in mind: IE. don't search for inspiration on Artstation/Behance but at the local library digging through books of the last century. We see the "art of prompting" right now, but to me this, feels quite limited. That's not the way I art-direct. I want to use drawing and paint-overs to communicate, I want to show reference images and point out the specifics I like, but let's see what the future brings. Edit: Oh, and polymodeling is already a hipster-thing, with all these PlayStation 1 styled games popping up right now ;)

I think it's a great tool for creative directors. I work as a 3D artist and as you described, it's kind of the same process you go through hashing out concepts in the studio. Think it will be a great tool to put in your project concept ideas and give you routes to develop from using tradition methods where you can build as you wish.

This was a fascinating episode. I use Dali2, an excellent tool for quickly imagining ideas, but it won't replace a 3D artist job soon. Perhaps the concept artist's job has gotten a lot easier and faster.

Loved this nerd rant. Hope you guys revisit the topic soon!

I was very nervous about the tech until this video - now im excited. Theres definitely going to be some affect for sure, but those that pay attention and use it to thier advantage will benefit greatly - its an entirely new medium, with entirely new functions. And hey if it mean i dont have to learn UV texturing, im all for it 😂

HI, interesting idea that poly modelling will become hipster activity. When I first started in 3d using infini - D it was very 'destructive' in that there was no way back after doing modelling things. This followed into c4d. Now with houdini, blender geo nodes and even c4d geo nodes everything is going procedural(I realise that existed before ,in probably max and houdini). Perhaps this AI will lead to some kind of hybrid - add midjourney node, then midjourney post process node etc.

very insightful discussion!

what I am grateful for about MidJourney is something that mostly everyone sees as a dysfunction of the AI tool. This is the approximation in the rendering of the forms and objects etc. The distortion seems to be an artifact of the inaccuracy in how it renders things specifically. Its has come a long way and for most people its at a level of magic in terms of understanding what is actually happening behind the scenes on the servers.. Maybe one day I will find a complete description of the logic behind the process but right now I am enjoying this mystification of the technology. Anyway my point is that because of this current inaccuracy in this Mid Journey tool, most open minded folks are still enthralled by the possibilities and as a saturation of these images continues this peculiar artifact will no doubt constitute an identifiable aesthetic that retains certain associations that are still evolving. I kind of see this as a kind of post-modern abstraction. And it makes abstraction even more acceptable for artistic pursuits especially in design. Right now in 3d animation and motion design there has been a push for photo realism and image fidelity which is needed in advertising and marketing for obvious reasons but I really believe the impact of this mid journey and Disco Diffusion will make a huge impact on loosening up of these all of these uptight creative whose ideas are all about photo realism and perfect little renders with perfect textures and perfect objects....i love seeing three eyeballs on a cat distorted unpredictable landscapes and mind bending perspectives. I love how the AI ignores the typical human's grounded perception and just fuses ideas together without any prejudice or predetermined expectations...I can see there will be a huge influence on the next waves of trends and styles just as the datamoshing etc had a moment of being popular... I think this may be far more reaching and impactful. Right now most people I know are experiementing and developing their approach as to how they will incorporate these tools into their workflow...

People do use artstation as prompts, but also existing artists. You can use it creatively, using for sketching. It can bring a lot of interesting things. BUT when random people generate work based on styles of other artists and then sell it ,that can easily become a IP issue.. Style of an artist could become the reason for IP lawsuits. Also artists should be able to submit a claim for excludimg their imtellictual propery from AI learning. There are responsibilities on all levels to make this end well. On the other side. this is the price you pay for not caring ebough about privacy and IP. Putting everything online thinking that Google has good intentions😎 So for artists long term. Don't put your art online. so an AI can't see it. Protecting your work will become extremely important to protect your IP. Show and sell your work usimg NFT' smart comtracts on the blockchain. use 4000 pixel artworks so the platform and collectors, fans know you are actually created something as an artist. When AI starts to generate flawless jazz band improvisation, i will be worried. And probably not here anymore 😆

I think the functionality of video input in projects like Disco Diffusion is really something to watch - I'm seeing people start to make really elaborate animations based on particle systems made in C4D/Houdini

I welcome our robot overlords. Having a blast with Midjourney and look forward to using it to help generate ideas in production.

Yes. It’s a fun tool to play with. And definitely going to use this as a starting point for future projects. But what I’m lacking is why is this the result? What are the decisions behind the different choices? With out this I can’t really tell if midjourney has solved any problem or not.

You are missing the obvious! > Resistance may be futile, but please hold the Robot Overlords off as long as possible! Love your work, and continued success!

If the AI could literally read the artist's mind and produce exactly what they have in mind, it would be entirely ideated/designed/created by the artist, only the technical aspect of transferring the idea from their mind to the paper/computer would be different. Now we can think of the current state of things as the AI being bad at reading the artist mind such that the artist has to explain everything through text, and some things are lost in translation. But this is a bridge that can be shortened from both sides, the AI being better at understanding you (and learning your individual, specific way to communicate), and the artist becoming better at explaining things to get the result they want.

Applied for midjourney beta. Not accepted yet... more time for Houdini!

You gave a prompt that was descriptive. What happens when your prompt is an emotion or interpersonal communication? An example would be "grieving over loss of a loved one"

Midjourney seems like a great tool for an early ideation phase.

I agree with most everything you guys say here. The artist's job is not dead yet. However... I can't help but thinking that if for example someone just needs something as simple as cover art for their album... will inputting their ideas into mid journey and slapping their title on top be more appealing to them than paying an artist $5000 or even $500. They may know its not as good, but considering they can ostensibly generate for free in 60 seconds... well they make the sacrifice?

love this!

@ 13:15 "The New York School."

What about economic factors? What if designers become dependent on this tool and then it becomes too expensive for them to use does that mean creativity will depend on who has the means to afford such an advanced tool?

Am I right that 1) Imagen has no public access at all, 2) DALL-E is waiting list only?

26:02 Nanite in Unreal Engine doesn't use polygons (it's a cluster of UVs) so we're already there 😄

How does this software decide on composition?

Can artists opt out of having their work scraped by these tools? If it recognizes artist names like James Jean, and is reminiscent of thousands of ArtStation images, then it means that aesthetically pleasing images are pivotal to the existence of this tool. Without good quality input images, the model doesn't have anything to mimic and build on. After seeing hundreds of the nightmare-inducing midjourney images on every social media platform, it's pretty apparent that meaningful and memorable images are going to become more important than ever before. Maybe some of those professors were onto something.

It's a great image generator.

Lots of talk about using this for ideation. Does that mean you accept machines are creative? And that they are better and faster at being creative?

I dont think either concept artist are doomed, but it is very important to notice change in supply and demand. With tools like this one concept artist is capable of doing many more jobs, so supply will driasticly change as demand will stay nearly the same

We are so stupid we are letting the machines have all the fun why? because there is a lack of creativity and a desire to produce a lot of content. why is there a lack of creativity because artist just spend all their time on social media no longer exploring the world the universe their minds. Only true artist that explore their own imagination and create things from scratch will get to explore and work directly and understand , transform and get to beauty, everyone else who lets a.i do the work are just selecting images or videos and praising the machine as incredible where it is really just a big banality

When people ask if this is a deathblow to designers, and a lot of content creators produce the same optimistic answer (which I find suspicious as a cynic who believes anyone making money from ad revenues has their vested interests) I prefer to say YES it is/will be a death blow to designers.

It is silly to think we as humans are anywhere close to replicating the intricate complexties of human mind, creativity and intuition. Everything meanwhile will find its place and reorder some supply demands. How our brain works and where ideas originate from will always be much ahead than the best of AI

Lol “public” sources. I’m sure there is a way to check that? Couldn’t be screen grabs right?

But on top of this mid journey has the sand to charge for their use. How about mid journey compensates the artists that contributed to its database?

I look forward to AI being used for texturing. So no need for UV's, would be amazing.

I see it as a meme generator

But would it come for us Houdini artists?

Do you think the cell phone rise and the amount of pictures that have been taken over the last 15 years has given AI a tremendous database of images to draw from and has allowed the AI to progress so rapidly along with the expanding power of GPUs.

Man. Just imagine when whole movies can be generated with AI.

Till people still have to do rotoscoping from time to time I feel pretty safe. All this huge AI developement (we should really use more machine learning name), and humans are still more accurate at finding edges while Elon promises self driving car every year. I think we are more far away from that moment than most people think.

i think if they make Ai help in retopology or uv mapping thats would be great cause those most dull ones for me as and artist lol

i think this can help in creating storyboards but concept art hight quality or realestic 3d rendering still far from mastering that

its funny because Midjourney is far from public.

"I know a genius in MidJourney", but you are going to keep her talent a secret?

Iove this discussion but there is so much whispering a mumbling.

In short- we need to think of ourselves as music producers not guitars players. Most music producers aren’t masters of anything - but know their way around many different instruments and genres. Further more a concept artist myself here’s my two cents - as a community we have to start becoming more aware of the ‘meta’ of how our art is interpreted in the modern world. Every medium of art has a certain ‘lean’ in the way that you can push and present the medium and have it be the most impressive. It’s going to be more important now as artists to begin to clearly advertise what we are ‘getting at’ with in a given style that we are using. This will make us much more getable to the masses and will help us always stand out from Ai. We need to emphasize the human elements of art.

"Death Blow to Designers" Why so negative and terminational? The goal of AI is to enhance human productivity. Instead of running teams of dozens of designers technology such as this will enable smaller studios and individuals to compete in a higher level and more organizations will be able to afford to have design embedded in-house. Also text communication with the AI isn't the only input possible, one can also design a solution and ask the machine to come up with alternatives, which today let's be frank, it's such a pain work to do.

My opinion is.... While technology and democratization of it is always a welcomed development for the community, ... what happens eventually is the amount of quality art starts to drop... Take a look at the current movies been released of recent and compare to back when CGI wasnt this accessible... it is easy to see there are plenty movies now with little quality of art but mostly just poor CGIs and no texture... Take for instance, so many movies were released last year, but from the top of my head now I can only remember DUNE as being exceptionally great... with all the technology enhancement at the disposal of the Fast and Furious Team, the last edition was no where as good as the old sequels... sometimes, tech enhancement kills craftsmanship when not well used ... tech is sposed to enhance art, good artists will find where it'd sit in their tool-box, and not drop their craftsmanship to accomodate new tech....