I mean I would HOPE people would need a real smoking gun to say that YOUR art is AI. I mean a lot of MFT is Anti-AI, so that would be random if it was made with AI.

Either way it's usually better to question, but not accuse when it comes to AI I think. Unless it is very much egregiously AI than it's better not to just throw accusations.

At least in this comic the vibe I get is just wanting to be a critic and not actually appreciating art though.

I can't imagine how frustrating this must be for artists.

My industry has been hit by this too, but for however much I want to believe that code is art, it will never be as personal and expressive as actual art. I really feel for you.

Love your work, btw.

Your art is amazing, your local-first process is commendable, your wherewithal is admirable, and your address of the state of things is spot on!

Absolutely love your stuff. I love that the lack of algorithm in the fediverse doesn't bubble up "engagement" for you, and that you know you're really valued!

I mean, by me.. Whoever I am.. BUT STILL! Love your work!

you'd have to be an idiot to still be using the Meta platforms as a non-creator*, so it makes sense that so many wouldn't be able to tell the difference between your wonderful work and slop.

* I understand creators and brands need to stay on there for their livelihoods. I wish they didn't, but I wouldn't call artists needing to make a living "idiots."

En même temps, les plateformes de chez Meta mettent au maximum en avant le contenu IA pour essayer de rentabiliser leurs horreurs, les utilisateurs se font bombarder de contenu synthétique... ils doivent être un peu rincés de tout ça.

Ici, tu postes un contenu IA, t'as la Fedi-police qui défonce ta porte d'entrée avec un bélier (un bélier furry, hein, faut pas oublier qu'on est sur le fediverse)

Et tant mieux !

I really like your artistic style.

Also: If someone sees your "making of" videos and all the Krita videos then it is quite hard to think that your stuff could be AI.

This is likely an effect of ranking algorithms on the Meta platform, which the fediverse doesn't use. The first controversial/shocking comment likely gets ranked up and pushes your post to be ranked up to create a spiral of negative engagement.

Human brains naturally engage more strongly with negative content, and so those algorithms are designed to exploit this to drive attention (and profits). Yes it's a very sad state of affairs. But the contrast with a non-algorithm-based platform is stark.

@dgouttegattat
A lot of technically-minded people use a lot of parentheticals to cover details, and em-dashes really help.

I suspect the AI uses them because it tends to random-walk around the point. It's bad at staying on topic. I always have the impression that it's weaving all over the road and bouncing off the guardrails.
@davidrevoy

@TerryHancock @dgouttegattat True, I wasn't familiar with em-dash before making the translation system for Pepper&Carrot, in French it's not common, we often use the : or ; to split sentences like that, and of course, yes, personally I use too many parenthesis (of course). But I also adopted it.

One of the AI effect I observed; now many French articles have this US/English way to construct sentences, with em-dash. And French tends to adopt it, so AI is shifting local culture more than we think.

@TerryHancock @dgouttegattat That's actually really interesting, and an effect I wouldn't have considered before. 🤔 I reckon it's a continuation the effect globalism has that blends borders where they touch.

In the case of English, I think there is a subtle difference between them, where one would be a more appropriate use than the other. But it's so subtle that most folks don't notice or care, and it's so ingrained into my instincts that I wouldn't know how to explain it anyway. 😅

it just hurts seeing this happen all the time

can't wait for the world to get better again

@mandrake Hn, I have been writing for the vast majority of my life & it's ridiculous to me that someone might demand a process vid for my writing. Absolutely not. You pay for that kinda shit on a Patreon tier.

And I get why you're doing it, & I'm not mad at you, personally, but I don't want writer process evidence normalized. Make them prove my lifelong cultivated skillset is LLMs. I won't change my software just for shitheels accusing me for their pleasure. That's harassment.

@Verati404
As David says, these machines were made to generate content that imitates the human form reasonably well.
That sucks for arts especially. But Pandora's box is open, and if we want to continue having publishers and readers, who care about supporting human created works, then we need to enable them to do so.
People who actively care about whether something with my name on it was actually made by me or is robot slop, are not my enemy. Quite the opposite.

I want publishers to be able to distinguish between human and LLM generated writing, so they can keep on supporting and publishing if no name newbies like me with confidence, and don't just give in to the slop wave. Using git for this is an extremely cheap, accessible and reasonably reliable way to do it. Even if I am never published, I hope they will continue to scrutinize works, because as a reader I'd like to continue spending my time reading peoples work, not robots.

It's not about my ego, it's about enabling an ecosystem I care about, to continue existing under radically changed circumstances.

@davidrevoy

@Purplegengar25 If it makes you feel better, I think the percentage itself is more of a confidence marker than an actual "this thing is 26% AI". Something with a low-but-not-zero probability can come from really basic standard typing habits and would be considered an unlikely positive, while something like 50-75% would still be a neutral uncertainty. When you start getting into the 90s, then you got something.

Unfortunately, the results readout can be very misleading if not understood well.

@Purplegengar25 Also--I ran out of space oops sorry--I find the more ethical ones run on an "innocent until proven guilty" thing and preferably don't speculate unless it's very confident. False negatives are way better and less damaging than false positives.

Course it might also depend on the model. My experience is mostly with SynthID, so... ymmv??

On the contrary. Thanks to you! It made me wonder how many images I’ve quickly closed just because they were “AI,” when they actually weren’t...

And on top of that, it seems you’re a FOSS advocate and a Krita user (and maybe Blender, if I understand correctly).

So clearly, I would have missed out on an opportunity, especially since I’m trying again to get into Krita, among other foss.

Et aparemment vous êtes français en plus. Haha.

@Cyber Yuki @David Revoy It gets extra tricky for these systems to reliably cite sources because of how jumbled around the training data gets. It's kind of like money laundering but for creative works.

If I'm not mistaken, there have been attempts to add citations to LLM output to try to improve their perceived trustworthiness, but it frequently just resulted in "hallucinated" citations.

I don't know how hard to lean on Hanlon's Razor here, but it does seem to be an issue that they don't have much incentive to correct.

@bassistance If you mean the comic, just check on my website: on the footer I always share a high resolution, the Krita source, the text separated on a transparent PNG or PDF and the comic with just the artworks too.

If it's about the fictive artwork the Gothic Sorceress Cepper is painting on this comic, you'll find it here ☺️ :

framapiaf.org/@davidrevoy/117043768651453481

David Revoy

2026-08-05 16:09:36

@ghosttie Her fictive painting? The largest resolution I have for this color sketch is peppercarrot.com/extras/temp/2026-08-04_gothic-sorceress-fictive-painting.jpg