My brushstrokes against AI-art

Published on

Picture above: concept art for episode 39 (click to enlarge).
A close-up of a piece of artwork showing Wasabi, but young.

A philosophical questioning

Recently, I've been browsing DeviantArt, ArtStation and other art platforms that have decided to empower AI artists, as if the recent and continuous enshittification of their platform wasn't strong enough.

But I have to say: AI-generated digital artworks are really improving and getting more and more things right: proportions, composition, even the number of fingers! They are getting closer and closer to complete imitations of real artists, and that is impressive in its own way.

But even if it sounds like I'm praising some aspect of AI-generated imagery, make no mistake: I'm still very much against it. I remind you that thousands of my artworks have been analysed over the last 20 years and are part of the LAION-5B database. All without my consent.

In this respect, I could ironically say that all generative AI stuff can be considered derivative of my work, even if only to a 0.00001% degree of derivation.

All of this is really energy consuming.

So this soup of feelings continues to influence my relationship to making new artwork, the way I approach the canvas and the way I paint. Because it challenges some deep underlying philosophical questions:

What is human in my work?
What is my signature?
What is "I"?

To be honest, I still don't have a clear answer to these questions, but my instincts are pushing me in one direction: to show more of my brushstrokes, my hand gestures, and to avoid smoothing things out.

Brushstrokes as a vector of identity

Let's start on the same page: I'm pretty sure that artificial intelligence art will eventually manage to parrot every artist's brushstrokes, if it hasn't already. But it could be very challenging: so far I see a lot of over-smoothed artwork, blurred or melted strokes, strokes with no start or end, and over-cleaned surfaces. Many results are devoid of personality and gesture.

That's probably why my taste for expressive brushstrokes, human gestures, has increased since my last art browsing session. Even my taste for artists who use the traditional is on the rise, and in my opinion it could even be a new trend.

But what about digital art? I think that we can still manage to express ourselves. With some artists, I'll always recognise their long brushstrokes, or their little 'loops', or their 'noise', or the way they accentuate their colours.

Could an artist's brushstrokes convey authenticity?

Maybe so. At least here is a hint where I can put my "human signature" in the most obvious way. Unfortunately, it is difficult for a painter to master the economy of the brush stroke. Resisting the urge to smooth a volume and 'let it go' is rarely my first reflex.

But I'm working on it, and things are starting to come together on this stylistic issue:

  • My recent pen training sketchbook really reminded me of how I loved to crosshatch, and how those crosshatches were personal gestures.
  • My new tablet helps me to transfer this kind of gesture to digital, thanks to the low parallax and fast response.
  • Also, during the recent exhibition of prints in giant format, I enjoyed seeing my brushwork magnified. Especially when I saw the scribbling and crosshatching done on the Dragonfish Restaurant, a 10 year old piece, printed in large (see zooms at 100%).

So that's what I'm working on at the moment, and what kind of thought chemistry is boiling in my brain.

Do you want to know the irony of all this? These developments towards a stronger personal style in my art, though I am pained to admit it, come from the pressure of the existence of AI-generated images.

Should I be grateful for that?

Picture above: concept art for episode 39 (click to enlarge).
Artwork showing Wasabi, but young.

Picture source here

License: "My brushstrokes against AI-art" by David Revoy − CC-BY 4.0
Tags:  #article   | Download: Markdown


link David Revoy Author, - Reply

⏳ Process:

197 ★

link Kathy Murdoch :pastille_lemon:   - Reply

Really interesting post! You're definitely not the first person I've seen suggesting that showing the human brush strokes will become a new trend.

link Olivier Saraja ☕️📚🦖🛸👻   - Reply

I was asking myself the same kind of question when 3d artwork started to arise vs 2d artwork in early 2000 (for me). At the time I was a poor artist with both, but I understand that for now on, good "prompters" will be able to make outstanding art within milliseconds. Luckily, these present days, there is still a societal resistance to IA-Generatives although, yes, this can be mesmerizing.

link Arne Babenhauserheide   - Reply

@oliviersaraja some context I heard about 3D becoming more dominant: one driving force seems to have been (as by some youtube video … sorry, no sufficiently reputable source) that the 2D artists were unionized in the US, the 3D artists were not, so 3D was cheaper since artists earned much less per hour of woork.

link Mahid (Moved to   - Reply

That last point about AI causing stronger styles is something I haven't really thought about but now can't stop thinking about

As a 3D artist, I can't really use stuff like brushstrokes, but I have noticed in the past year I've been moving towards more complex texturing and materials (which is unusual for the specific subject matter I work with)

I'm interested to see what larger scale effects we'll start to see over the next couple of years

link David Revoy Author, - Reply

@standingpad Same, I'm very curious.
Very interesting to read your feedback with the 3D point of view.

link Felix   - Reply

Good write-up. As to your concluding question: I don't know, but it's a glorious portrait. That's the most important thing.

link David Revoy Author, - Reply

@felix 😊 Thank you.

link ammoniumperchlorate   - Reply

je crois qu'il y a une autre différence très importante : Tandis que le CGI statistique / "arte IA" ne fait que reproduire la distribution de probabilité de ses données d'entraînement, les artistes humains sont capables de réfléchir sur l'esthétique de leur style et le faire évoluer. Par exemple, le style animé japonais a beaucoup changé depuis les années 1960s où il a été conçu - si on avait utilisé du CGI statistique pour produire des nouveaux animés a cette époque, on n'aurait

link ammoniumperchlorate   - Reply

jamais eu la richesse des sous-types du style anime (je pense aux différences entre Astro Boy, Mononoke, Chihiro, Assasination Classroom etc.)

Et puisque vous vous avez posé la question que faire pour vous distinguer de l'arte IA, je voudrais vous dire quelque chose: Votre style est déjà unique et magnifique :) J'adore vraiment comment vous avez pris le style anime japonais comme base et ajouté des détail comme on ne les voit jamais dans l'animé classique. C'est déjà quelque chose

link ammoniumperchlorate   - Reply

que l'"IA" d'aujourd'hui ne peut pas faire du tout (et desolée si je parle du non-sens d'un point de vue artistique, je suis physicienne 😅)

link David Revoy Author, - Reply

@ammoniumperchlorate :blobaww: Merci pour les mots sympas sur mon style.

C'est sure que je suis également très curieux sur ce que l'intelligence collective des artistes vivants maintenant va produire en réaction à tout cela sur une échelle de 20 à 30 ans d'ici là.

link Martelo Schwarz   - Reply

Thanks for the insightful text. Really enjoyed reading it.

link rfnix   - Reply

that article reminds me that I've heard similar things from photographers who show disdain to the heavily denoised and flat images coming from smartphones and say "noise is good, show more of that!". Which is I guess why some go for analog film b/c of that distinct structure!

Don't you think the next generation of AI art will have a "draw like David" mode? (and I'm not saying it like it's a good thing...)

link David Revoy Author, - Reply

@rfnix Interesting for photographer. Yes, I saw a trend toward analog film and also old glitchy or compressed low resolution first gen of digital camera.

You can already ask Stable Diffusion (or any using the LAION-5B database) "in David Revoy style". But the output will not really works as good as for other more famous artist 🙃.

A real "mod" like a special training? Maybe. If what I do will be of value at one point, I'm confident someone will do it (even if I disagree with it).

link Guillaume ☭:yunohost: :mageia:   - Reply

merci ça me fait une belle photo de verrouillage de téléphone

link David Revoy Author, - Reply

@Guygoye Chouette, ça me fait plaisir que çe soit utilisé. La 'full resolution' est ici :

link Trygve Kalland 🇺🇦   - Reply

“These developments towards a stronger personal style in my art, though I am pained to admit it, come from the pressure of the existence of AI-generated images.

Should I be grateful for that?”

Be grateful you’ve risen to the challenge. No need to be grateful to the threat to your livelihood.

link David Revoy Author, - Reply

@trygvekalland Thank you. That's indeed a better way to look at it. 👍

2 ★

link Ricky Romero   - Reply

This is a gorgeous painting and an interesting take.

I don’t know that these “prompt engineers” are unable to produce this more crafted art style using AI. (If not now, it’s likely only a couple of years away…) The question is, will they ever have the taste to perceive it as an asset rather than a flaw? 😔

link Codrus 🇺🇲   - Reply

I've taken a number of Art History courses in college so far. It really is a historical trend for artists to become more "painterly" as they learn more. Titian started with perfectly smooth paintings, and by the end of his (very long) life he had brushstrokes everywhere. You're in good company!

link David Revoy Author, - Reply

@codrusofathens Haha, thanks for mentioning this. Yes, it feels like when the classic painters faced photography and new art movement like expressionism, impressionism, etc... started to exist. Maybe I'm just revisiting this unconsciously, 150 years later.

2 ★

link Mobius Goddess   - Reply

I think that AI gen will have to be fought. Politically/legally, and then by using tools that try to protect your style against counterfeit machine like Glaze/Nightshade.

link Mahid (Moved to   - Reply

@foxysen I find projects like Glaze a band-aid, not a proper solution

The way I see it, the proper solution would be to let us artists have control over whether our art is used in datasets. Speaking as a 3D artist, AI can be a powerful tool (like creating textures for 3D models), but that lack of control over datasets is really my main issue with AI

link ratfactor   - Reply

I've been having very similar thoughts lately. Distinctive brush work and traditional materials that show the human process are becoming important.

I feel truly awful for digital artists whose work looks most resembles the generated stuff. (Stanley "Artgerm" Lau comes to mind. Superficially, the machine stuff is so damn similar to his beautiful work and I assume that's a direct result of both his popularity and the flawlessly smooth quality of his style.)

link F4GRX Sébastien   - Reply

mais c'est de la folie ta maitrise. wow.

link David Revoy Author, - Reply

@f4grx Merci.

link LonM   - Reply

I was thinking this recently with respect to LLMs and text. It can push you to refine what you're doing to be a bit more different and diverse. But there are limits. At what point does your approach, changed to be distinct from AI, stop being your own approach?

link David Revoy Author, - Reply

@LonM It's an interesting question. I'll probably see that with a bit of temporal distance, when I'll look back at what I'm doing now, if I feel like "oh my... I was really exploring edges of my own style". Because I'm confident time will always 'recenter' my style where it should be. I just hope that I'm closer to it, and not taking distance with it. 🙂

link Valentin Petzel   - Reply

I do believe that there is more to art than what AI can easily replace. To be fair - style is always replicable, and often not that specific. What is the real difference between AI copying some artist and some student copying some artist? Art does have a dimension that is completely abstract from the thing in question. A painting in the style of the Mona Lisa still is not the same as the Mona Lisa. Art is more than just production of repetitive stuff. ...

link Valentin Petzel   - Reply

... And surely there will be artists who make art by using ML technology. But this does not change that art is more than just the final product.

For example why do I pay for Pepper and Carrot? After all there is already so much stuff readily available online. Well — for one thing because my (by Friday) 5 yo child is a big fan, so I asked him if he wanted to support the comic. But it is also because I care about the process behind it — not the result.

link David Revoy Author, - Reply

@valentin_petzel I agree.

And thank him very much! Here is a tiny doodle for his birthday:

Carrot blowing 5 candles 6 ★

link Valentin Petzel   - Reply

Thank you very much, I'm sure he'll love it!

link Valentin Petzel   - Reply

🖼️ c2cbfb9d2a8c8cce.jpg 

link David Revoy Author, - Reply

@valentin_petzel 😍 😍 😍 Bravo!

link John Colagioia   - Reply

My opinion shouldn't carry much weight, but I wonder if the real solution to AI art is mostly to just let it run its course. After all, it lacks something much more important than brush strokes: Intended audiences and fitness for a purpose.

An AI might make something "close enough" for a project, but the person making that compromise would probably have just used clip art, in past decades. And it can't do better, because explaining all the context would grow prohibitively expensive.

link cliffp   - Reply

Every single AI generated image uses learning images fed to the various AI machines. All AI art is based on human art. Every single dot of it.

Without the input of human art, AI art will stagnate. Without human input, from the front end by instructions, or in the back end by scanned and analysed art, AI art is going nowhere.

Treat it like a complex brush, and remember that AI is by nature derivative. It's not the brush strokes that make yours original but the input of your brain.

link Júlio Gardona   - Reply

AI art is cool when you need references. Thei dont have expression. All ai art seems generic to me. Thei are the same noses, the same eyes, the same lightning and colors.

link Ray Of Sunlight   - Reply

Funny thing: AI-Generated Art has no copyright because it only protects HUMAN work, not AI-Generated Work, also, yeah, relying on AI Art it's just lazy.

link Mahid (Moved to   - Reply

@Ray_Of_Sunlight in the US at least, most countries have yet to weigh in on AI art and copyright

link Shark Attak   - Reply

I can't say I have the time and eye to discern AI art, so I take a simple approach: If it has AI on, I don't watch. Not interested.
Not because it can't be good, but cause it will surely exploited to pay people less, so..

link Arne Babenhauserheide   - Reply

I like seeing more of the artistic side, but keep in mind that also for the craft side, we’re far from being at the end of the road.

There will still be people pushing the boundaries of how to make a canvas appear real — be it digital or analog — and AI can only make their techniques available to many more, but not invent new ones (at least for now).

link Kiran 🏳️‍⚧️   - Reply

wowow!!!! it's always mesmerising to watch your drawing process!!! The drawing looks soo good!!

link Tyler Mumford   - Reply

My goodness, that's awe-inspiring.

link Lloyd Weehuizen   - Reply

this is amazing, thank you for sharing! My 9 year old son would like to know how long this took you?

link Thomas Frans 🇺🇦   - Reply

It's always fun watching a black and white shape turn into artwork.

link Starbeamrainbowlabs   - Reply

This is so cool! The trick there of getting the shape right in black and white first before adding colour through filter(?) layers is such a neat idea!

Drawing people is very difficult for me atm, so always interested in learning new tricks.

Great artwork - thanks so much for sharing with all of us <3

link David Revoy Author, - Reply

@sbrl Thanks! Yes, the workflow is really similar to the one I introduced into this old tutorial if it can help.

link I'm listening   - Reply

Very cool!

link felix (grayscale) 🐺   - Reply

also note, this era of generative-ML art is fundamentally bad at 3d. they fake 3d with 2d patterns that look good locally, but don't make sense in an actual scene. this becomes noticeable as incoherent lighting/shadows, inconsistent vanishing points, etc.

these are things that human artists get wrong too, so it's not a super reliable indicator, but generative-ML is often weirdly very good at some things and very bad at other things, in a way that humans aren't

link felix (grayscale) 🐺   - Reply

I think it's extremely unlikely that generative-ML will get better at 3d without using a different NN architecture. LLM architecture is designed to look at linear sequences, and they struggle badly if you ask them to do anything 2d like ascii art. Stable diffusion and other art generators are designed to look at 2d images, and there isn't an easy way for that architecture to learn 3d representation. (real brains have networks specifically for 3d, and people struggle with 4d)

link   - Reply

The way I see it there’s two aspects of this. The continued breakdown of capitalism (the move towards post scarcity) and the extreme hoarding of wealth. This scares us, as we still need money to live.

Then it’s the human aspect. Are we becoming obsolete somehow? Are we being copied?

In the first case it’s political. Pretty sure it’s easy to figure out how I lean there.

But in the second case I think humans in the end crave a human connection. Even if a computer could churn out music and art by itself, it would lack a human connection. Someone writing a prompt isn’t enough.

Art is an exchange between humans ultimately, imho. Viewing your art (since I’m replying to you) means more to me because I know you (as in I read your posts, have seen videos, follow you. I’m not stalking:)).

Imho I think it’s valuable for artists to remember that.

Using someone’s work to train an “AI” without their consent is imho unethical. Consent works on many levels.

With that said, what’s your thoughts of having your own personal AI assistant that is exclusively fed by you? Whatever form that would take. A computer, tablet, and software could be seen as an assistant after all.

Just some thoughts. Even if I’m not an artist (can’t draw a stick figure that’s any good) I like reading your thoughts on the subject. Again the human aspect.

Looking forward to the next chapter. I swear she had a bit of a manga inspiration when I watched the sped up video:)

(Original message has been truncated: read the complete original message here.)

link stefan   - Reply

This is a very interesting. It does also give some food for thought as well. You know I am a big fan of Light Novel and Manga. There is so many different artists and some different styles. A Light Novel is mostly literary but features some illustrations that feature some scenes here and there.

For some modern stuff the illustrations in books, manga and anime started to look very same-y to me. This likely is not the case, but there is that modern style especially in anime that makes one show kinda look like the next.

link stefan   - Reply

When you hit send way to early... Anyway back to the point.

Maybe you know what I am talking about. But that "normal" style makes stuff outside the box really stand out to me and watch in awe, when an illustrator puts some really unique work.

Two illustrators that really show this are so-bin (Illustrator for Overlord and Blade & Bastard) and lansane (Illustrator for Min-Maxing my TRPG Build in another world)

so-bins stuff looks more like a painting than an illustration for a Light Novel especially the colored ones are just gorgeous to look at.

lansane is interesting as well festuring a pretty distict (to me) artstyle sometimes with very thick almost sketch-like lines combined with the use of raster foils typically used in manga.

I can post some examples as well later if you like a look at what I try to describe by words here.

link Lost Prophet 🇵🇸   - Reply

The flowing of the fabric is so beautiful!

link raghukamath Krita team, - Reply

This is awesome. I think in a way a new art movement will arise from this chaos, just like how the earlier art movements were a reaction to some events and trends in past.

link Pool dessins   - Reply

Salut David, intéressante et intense réflexion. Comme je commence à me faire vieux, ta réaction me fait penser à celle que j'avais eu en redécouvrant les Marvel's Comics et à la déception ressentie face aux dessins par trop "léchés" et "lisses" de cette nouvelle mouture par rapport au "Strange" de mon adolescence. Les outils digitaux ont certes aidé à la production d'images toujours plus complexes et "pétantes" mais ils ont par ailleurs formaté l'oeil de celui qui regarde.

link Pool dessins   - Reply

Il est de plus en plus difficile de percevoir l'âme, au sens poétique du terme, d'un artiste derrière l'œuvre tant nos yeux sont de plus en plus bombardés d'images policées et "propres". L'IA nous relègue au rang de tâcherons, superviseurs désabusés du travail de la machine. Pour moi la vraie question est : quelle est notre valeur ajoutée ? Quelle est notre poésie ? Devons-nous nous soumettre à cette vision uniformisée et standardisée ou, comme chaque fois, nous rebeller ? 🙂👍

link DECAY   - Reply

This is not specific to AI, since 2000 there is a large standardisation movement in arts in general. It's really noticeable in music.

My position is if you want something that sounds/looks like everything else you should definitly use an AI or use the work of someone who likes to make standardized art.
I don't really like to make standardized music so I mostly don't do any. And I think it's fine, there is room for any kind of art in the world.

link DECAY   - Reply

I really don't care about AI, because I don't feel endangered by it.

The problem isn't AI, it's artists feeling endangered. They feel that way sometimes because of pride, but mostly for financial reasons... And money is an all other kind of debat.

link Brian   - Reply

I love visible strokes, so I'm happy to see you leaving them, but in a similar way to @cliffp and @yon comments, I think the real uniqueness comes from the simple fact that you are a person like us, who has similar struggles, your own experiences and views on the world, which all play a part in how and what you draw, our personal life experiences can't be copied by a machine (or human for that matter). 1/3

link Ryuno-Ki   - Reply

Thanks for writing up your thoughts.

Now I'm not a digital artist but if you allow me to extend it to creatives, I feel I can contribute a little to the conversation.

In my opinion, there will be a countercurrent to AI generated work: the human factor.

When I read up about design, I've learned that the industry is pretty individualistic: the art is the identifying factor of the artist.

This could weaken you.

Let me explain:

link Ryuno-Ki   - Reply

There are industries that a knit tight. This year I've learned about „Silent Key“ databases.
That's a community of amateur radio operators. Morse. Each person had their significant „key“: the way they handled the device. So much so that the community could tell who's on the other end by listening to the signal.

I compare that to your brush strokes.

The Silent Key is a database of operators who passed away. Their key will be forever silent.

link Ryuno-Ki   - Reply

I can imagine that if artists have each others back they could raise the tide for everyone.

Right now there is an imbalance between supply and demand. There are way more people who would like to use art and way too few people who could provide supply.

AI is „cheap“ (and unethical, no question).

So imagine I would approach you with a commission to something 3D like, you would not be the right address. But you might refer me to someone.

Add a small commission and it's win-win-win.

link Ryuno-Ki   - Reply

Because to me, it's hard to tell the quality.

What makes a work original?

How do I even know what I want?

I haven't used any of the generative AI services, but judging from description they present you with a set of suggestions and allow you to refine the prompt.

Kind of self-service consulting.

An empathetic human might be more efficient. But human labour is expensive (rightfully so).

link Ryuno-Ki   - Reply

It could be the other way round as well:

If someone approaches you with a gig that involves a cameo appearance to an existing work, there would be a phase to understand what is the expected outcome.

I have read a design agency blogging about using AI to „be wrong faster“.

Traditionally I would imagine the exploration would use sketches or perhaps storybooks.

But I can understand if a client would like to have more agency here.
After all the image is in their mind.

link Ryuno-Ki   - Reply

From how I plan to handle the AI situation (since code is affected as well):

Build relationships and rapport to understand who's good in what and what they prefer to work on.

Then refer to them when I receive offers I can't or won't handle. And hope for reciprocal behaviour.

My „brush strokes“ are going to be comments in the code and documentation explaining the thought process and discarded alternatives.

link Cute_Slime   - Reply

I love to see your evolution on this subject, AI can be terrifying, but at the same time is unavoidable.
even if you are not for it, you clearly grow from it and didn't let IA push you down.
for all artist that fear IA, don't worry, you will always be better than some lambda using IA.
I have seen what a true artist can do with IA, and it's really amazing, the fact that he is an artist play a major role on the result, IA is just a new tool, powerful indeed, but a tool.

Post a reply

The comments are synchronised every 12h with the replies to this post on Mastodon:

How to use this? (click here to unfold)
Open a new Mastodon account on the server of your choice. Then, Copy/Paste the adress above in your Mastodon 'Search' field. The post will appear and you'll be able to fully interact with it. You'll have full control of your posts: edit, remove, etc. After that, your message will appear here.

Just please note that it may take up to 12 hours for your changes to be reflected here.