BlackJackMacBastard

Hey gang! My generator of preference has been on the fritz for a few days, so no new generations from me until they get it sorted without it costing me precious precious minutes waiting on error messages. :)

But in the mean time I’ve run into something a few times that feels super suspect/scam-y. Twice now I’ve had folks using very similar language tell me they wanted to use one of my generations as the basis for a commissioned piece of art, and wanted to pay me. They asked for my cell phone, or contact via discord, or some other messaging app.

I do AI generated fetish images of copyrighted characters, and I think my ability to claim a copyright on *anything* I’ve ever generated is legally fraught on multiple levels. I absolutely don’t believe someone is going to toss me 500 bucks for the right to use a sexy Shego bondage image for a commissioned art piece, and if they are I think there are ways to convince me it’s legit (like have a DA account more than a few days old, or the ability to direct me to a portfolio of similar works, or literally anything).

My assumption is the scam works on some sort of reverse charge grift. I don’t know, and I don’t want to find out and accidentally get snookered. But I wanted to toss this out in case anyone else is coming across the same.

Remember friends, watch your asses out there. I’m just trying to get my kink on out here and am not looking to make a career out of smut; I have neither the time nor the interest to try to spin up the sort of assembly line that leads to real income out of AI porn. But if this *is* your livelihood, don’t get suckered in without at least some kind of due diligence to make sure the person you’re talking to isn’t selling you a load of shenanigans.

I’ve gone through my galleries and done some re-organization, which included taking any content that I think is riding the line of DA’s policies and tossing it behind the paywall. This means any content where a damsel is being penetrated, where she’s playing with an explicit sex toy, or where genitals are being touched.

On a practical level this means I have two new folders: Deviant Devices, where all things dildos and vibrators can go, and Foot Slaves, where any image depicting a footjob will go. Those images will still be in their original gallery, so if it’s part of a series or sequence, it’s still there, but now it’s greyed out behind the $1 paywall.

Going forward everything will be posted/organized along the same themes. Big sorry to anyone who had a favorite yanked out from being visible to the public to now requiring a subscription; but the more visibility these things get, the more paranoid I am about making sure I’m in compliance with DA’s rules to avoid headache and hassle.

I have this immense backlog of stuff I’ve generated that is too spicy for DA. If I’m being fair, some of the stuff I’ve posted might already be riding that edge. I don’t want to have to fuck around with getting my account actioned and my stuff taken down, so the most expedient thing is to create a minimalistic paywall.

This has the added benefit of letting me clear out my works-in-progress folder; it was starting to get very chonky and I’ve been neglecting the hundred or so generations waiting there to be polished or discarded. Having a paywall comes with a few downsides; for one thing I need to update to a Core membership, but that’s reasonably inexpensive. For another it definitely reduces the exposure of the stuff that’s in the private folder, but I suppose that’s the point. What you can expect: smut, smut, and more smut. I have a fair amount of material to trickle into the Private Stash, so there should be at least several updates a week until the backlog gets cleared out. I intend to keep the price at the minimal 1 buck. Subscribing gives you access to everything in the folder, always. I don’t want to have to goof around with time-gating material. The whole point of this is to be compliant with DA’s standards. I’ll be happy if this core subscription pays for itself, but I got no expectations there. Please feel free to drop a dollar, check it out, download everything, then come back next year. Or if a buck doesn’t mean much to you, stick around and get the steady flow of new extra-racey stuff. It’s all on you, heroes. I’ll continue to post to my regular galleries and mostly focus stuff that I think is extra mature, or that I want to be extra private into the Secret Stash. Some of my currently published bits may be upgraded into the stash as well, if I think it’s riding the edge of DA policies. That’s it! Now back to your regularly scheduled kink.

A question I’ve been asked a few times:

How long does it take to generate an image like (example).

It’s really wildly varied. Sometimes the generator puts something together that just seems perfect. Sometimes it’s like a tug of war where you’re trying to get a key detail without losing quality elsewhere. But I have a sort of system in place for when I’m generating specific images, either for a request, or just noodling around.

First I do the prompts. Each AI has its own system for these, and I think reading through tutorials and documentation on the generator you use is worthwhile. There’s a lot of promp-based tricks that help the generator churn out an image that both resembles the goal (a character in a situation, whatever) and is aesthetically pleasing. The first part is pretty easy. It’s the second that’s hard.

I’ll generate a series of images using the same prompt, just randomly rolling the dice to see what comes up. There might be a trend of results that leads me to edit my prompts a bit. If I see an image that I really like, an amazing pose, expression, whatever, I’ll save the image for later. This does two things:

First, it saves the image, and I can import that image back into the generator later, which lets me recapture the prompts and seed that generated it. This is awesome. It means so long as I keep the original unaltered image I can re-generate it with no loss of quality, if I wanted to work on it more.

Secondly it lets me consider the image and whether it’s worth saving. When I’m done generating, I’ll go through whatever images resulted and I’ll start sorting them:

1) Images not worth keeping. The details aren’t great, or there’s an anatomical flaw that would be too big a pain to fix, or something else just doesn’t work. It’s interesting to me to note that as I get more experience with generating these things, my standards for what is worth keeping have increased a lot. There’s stuff in my gallery I generated a month ago that I wouldn’t bother to keep today.

2) Images that need more generation. The AI I use lets me do some clever pinpoint adjustments on images. I can load an image with details I like, and mark a masking layer. Now the prompts will only change the areas of the image I’ve marked. This has been a go-to for refining images, but there’s a cost. Each new iteration of the image is slightly different, even in the unmasked portions. And so if I mark up half an image and regenerate it, then take that new image and mark up a quarter, and regenerate it, then and eighth, and so on, I’m zeroing in on the perfect results I want. But I’m no longer looking at my original image plus alterations. I’m looking at an image four generations removed from that.

Sometimes this is fine and either no detail is lost or it’s minimal. Sometimes it really changes how much I like the base image, like when line and shading start looking too rough.

3) Images that need hand-editing. Sometimes I can go right from generation to this step, and sometimes I need to do the steps above to refine the image and polish off the last stubborn details by hand. As a rule I only do this with relatively small edits, since I’m not trying to invest in a pro set-up. I literally use an iPad, the build in image editing software, and an Apple Pencil. I’m not a particularly good artist, but I’m patient and have a pretty good eye for color matching.

This step is why all my images (or maybe almost all) have the right number of toes and fingers. It’s a small thing, but those tiny details are what flag AI generated images in our brains, and cause uncanny valley. I think the impact to quality is really high when you do those touch-ups.

The generator I use also lets me import a base image, and the generator will try to iterate based on that image. This can be really useful when I do a hand-edit that stands out too much. Re-generating the image with the edit in it can blend it in; but again, at the risk of changing or losing details elsewhere.

4) Images that require nothing and are great out of the box. Rare, but it happens! Sometimes the generator will just serve up solid gold.

Once I’ve sorted the images I spend some time goofing around with them. Anything that requires re-generation I import back into the generator and try to refine. I’ve spent time on images that I’ve later discarded as unworkable. The goal is to get the image into a state where it’s either good as is, or I can finish any edits myself.

When I have final images I look at them again, now for an eye for anything that feels like it’s not up to quality, or anything too repetitive. Repetition is hard to avoid in AI generation since it’s an inherently iterative process, but generally I want any two images in my gallery to either be distinct, or for the difference to serve a clear purpose (an alt pic). I don’t want to do ten images of the same model in the same situation in the same pose, all with only small variations; I’d rather have two or three really good options that are different enough as to serve slightly different tastes.

Mileage varies here. What I consider sufficiently varied may not be what someone else does, and ultimately everything I generate explicitly serves my specific tastes.

The last step for me is if I want to arrange the results into some sort of series on my DA page. That can be fun, and I think it’s a neat way to organize images by theme. I experimented with writing a little mini-story under one series and will probably do that now and again.

Sometimes I have images that I really like, but that I’ve done to death. A series is a great place to put those where it doesn’t feel like more of the same. Right now I have about thirty perfectly good images waiting to be posted that just require me to decide how they’ll be posted.

From start to finish a single image can take as little as five minutes. That’s if the generator nails it right away. A longer, more complex image can take an hour or more of repetitive generation, followed by an hour or so of manual editing. The absolute longest time investments I’ve had on a single image is probably not more than two hours, but because I do them it batches it’s not super precise.

Either way it’s an amazingly small amount of time for a thing that would take a real artists hours or days to complete.

I’ve been asked a couple times what I use to generate images.

I‘ve used NovelAI for a while for generative text, which I’ve experimented with for table top games and writing stories. Like the name suggests, their primary focus is generative text, but they had an image generator I toyed with a bit. When I first played with it everything I tried to produce was Kronenberg nightmare fuel, and so I left it alone.

Then about a month ago they updated their image generation model based on StabilityAI’s SDXL model, and suddenly it was generating reliably coherent content. This model is trained with a focus on Anime style art, which is why the stuff I generate has that animated cel-shaded look.

When I generate stuff I start with a pile of prompts, and take my sweet time generating images, then trying to refine things that look like they have potential. Almost everything I posted to my gallery required some very minor degree of touch-up. For this I use the image editing built into NovelAI, or the built-in image editor on my iPad, with an Apple Pencil to scrub out bad lines and get rid of extra toes.

So that’s my non-rigorous non-scientific process in a nutshell: generate via NovelAI, refine and regenerate as needed, then edit on my iPad using extremely basic tools.

The most shocking thing to me is the speed with which material can be created. I could probably make a hundred images a day at about the same quality as what I already have, if I was willing to have a lot of them be iterations on one another. Even now there’s a little bit of that, but generally speaking I’m trying to produce the images that *I* want to see, and that means I will tend not to make a series of images that look near-identical to one another because that’s less fun. But it does take more time since the prompts have to be reset or adjusted. I also don’t mass-produce the generations because it’s more expensive to do it that way. NovelAI has a subscription model that lets me generate infinite images of a certain quality, but only one at a time. But that’s fine because I’m not super concerned with volume.

Anyhow, that’s the deal. I recommend NovelAI, especially for text. It’s a lot of fun to write stories with the illusion of interactivity. For image generation I suspect there’s better stuff out there that can produce more varied styles, but I’m gonna dance with the girl that brought me.