I'm not really into blog, twitter or social media. Instead I will use Rentry. It allow me to have a more casual presentation than bio from websites. Excuse my poor english, I'm only using ChatGPT for grammar correction as I want to keep my own tone. Sorry in advance for the simple phrasing and lack of words.


Who are you?

gerhard96, male, 35+ y/o, web developer.

And you do AI?

Yep. Since about March 2024. I started on websites like Yodayo, Pixai and then quickly switched to NovelAI for generation and Pixiv & DeviantArt for publication. I'm not a pro or an artist in any case, I just generate images and animations in my free time.

AI sucks!

Maybe. I'm not here to fight. The only thing I would say is that it took me a good amount of time and human decisions to generate something really fine.

Where can I check your "works" ?

On Pixiv, DeviantArt and Patreon. You can find all my links below.

Patreon ! Ha ! I knew it. You're one of those leecher looking for easy money.

Nope, everything I publish is completely free. Even my Patreon. I post the same content at the same tile on all three sites. To give a free access to Patreon I use a Gift Link.

Why would you do that? That's suspicious...

Well, to be honest, I tried to monetize my content for about two months. It was an Early Access tier on Patreon and, while it wasn’t very popular, I could totally have kept it going. But juggling early access and public releases became too much hassle. And making paid content slightly changed my perception of what I was doing. It was no longer just a hobby and started to feel a bit like work. Like telling myself “Ok, now I must not skip a publishing day”.

In the end, trying the paid route just… well, lowered my enjoyment of making AI content. So I gave up on that.

Ok. Give me your prompt plz ! I want to generate these sweaty feet too.

A lot of users ask me about this. Here is a good place to explain a bit about my workflow, and why prompt sharing is almost useless.

I do not simply “prompt and roll” when generating an image. I'm using features such as Base Image, Vibe Transfer, Character Reference and Inpainting. Basically, each time I get a nice image from AI, I use it as a 'checkpoint'. Then maybe I'll reimport this image as Base Image, changing the strength redraw factor and the prompt. Or maybe I'll use Inpaiting, changing only a part of the image while adjusting the prompt again. Or maybe I'll add or remove other images as Vibes Transfer, chaning the global style output. More than generating I would compare this workflow to "crafting" an image.

Each of those steps acts like a new image and "reset" the previous prompt and parameters. That's what I mean by checkpoint. Basically, to be able to regenerate the same image using it prompt and parameters, you will also need the previous checkpoint image, which itself would need the previous checkpoint image... I use these advanced functions maybe 50, 100 times during the process of crafting the image. Obviously I don't keep all these checkpoint images. So even I would not be able to regenerate my images from scratch. The only thing I can share is the last prompt used, which can greatly vary from the starting prompt.

A bit of word about Inpainting as I see a lot of AI users totally miss it: It's super useful. Probably the most useful tool in AI. Let me explain : Inpainting allows the AI to redraw only a specific part of an image, with the prompt applying solely to that section. This enables you to correct details, redesign elements, adjust facial expressions, and refine the composition — each adjustment having a totally new prompt.

Maybe one day I'll make a full tutorial. And, I'm not bullshitting you, with Inpainting I'm pretty sure I could generate a really nice image using only 1 word prompt.

You have text on some of your images. Is that AI generated too?

Nope. Once the image on NovelAI is finished, I use RealESRGan for upscaling it to about 4K and then I reimport it into a Photoshop-like software. From here I'm adding text, font-shadows and a signature. Then, I scale the image back to it original size (25%) or sometimes a bit larger (29% or 33.33%).

Before you ask : Yes, i wrote the text myself. I only ask ChatGPT for spelling correction and also for a full japanese translation (as I post a JP version on Pixiv and don't read Japanese).

And how do you make your video?

More than a video I would say they're animated images. Anyway, I use Tensor Art for that. I import one of my own images as image2video, choose WAN2.2-i2v-a14b-low-noise as Model, use the auto prompt feature (which is pretty good) and start generating some low quality 8 steps 3-sec shots. I refine the prompt until I get an OK render. Then I start to generate more lengthy (5/8 second) HQ (13 steps) shots. That's because — contrary to NovelAI where I can generate unlimited images — with Tensor Art each video costs credits. So I want to be sure the prompt is accurate before generating more credit-consuming stuff.

Once I get enough good shots, I use DaVinci Resolve for video editing. Here comes a lengthy work of choosing the right clip, adding another one at the right moment with the right transition to get the most seamless render possible. Basically it's video editing skills.

Ok and can I req-

Wait! It's not over. Once i get the MP4, my goal is to make a HQ GIF because Pixiv doesn't allow MP4. So I use Shutter Encoder, which extracts all the frames in the video as PNG (scale set to AI animation x4). And finally, I use Gifski to reconstruct all the PNGs into a GIF. Since Pixiv allows up to 15 MB GIF it's a trial and error, changing the resolution to get the highest-quality GIF possible.

That's great! Can I request a work from you?

Sure. Send me a message on DeviantArt or Pixiv explaining what you have in mind. If it fits my taste I will do your request freely. But note that I usually get a lot of requests in queue usually, and I do them on a first-come first-serve basis. This may mean a long delay before it's your turn. I don't do shota / loli / guro.

Can I use your work ?

Sure! The only thing I ask is to be credited. I can even give you a version of my images without signature (as I get stoleposted too much, I ended up adding a signature on each of my works).

Can I be your dog ? I will be a good dog. Woof!

Sorry, but I’m not into RP at all. My English isn’t good enough to write proper sentences, that’s why most of my images only have short punchy captions instead of long dialogues. And anyway, erotic roleplay has never really interested me.

Images are nice but what about a real experience? I’m allowing you to be my slave! Come to my Discord.

Thanks, but it’s the same answer as above: I’m not interested. I’m only here to share kinky content with everyone and do some of the requests I get. I’m not looking for any real-life interaction.


I publish 5 times per week, usually Tuesday Thursday Friday Saturday Sunday at 3PM UTC

https://www.deviantart.com/gerhard96        Posting daily content for free (medium censorship)
https://www.pixiv.net/en/users/105394842  Posting daily content for free (light censorship)
https://www.patreon.com/gerhard96           Posting daily content for free (no censorship, use the Gift Link below to access it)
https://linktr.ee/gerhard.96                           My linktree
https://www.patreon.com/gerhard96/redeem/4CE06  Gift link to access my Patreon freely (1year access, no CC required, no auto-renew)

Contact

I don't have any social media, discord, or mail. If you want to contact me, please use the DeviantArt, Pixiv or Patreon chat

Costs

I think it’s interesting to be transparent about how much it costs me to run AI-generated content. (Pixiv and DeviantArt are mostly for analytics and visibility.)

Pixiv Premium → $3.5
DeviantArt Core Plus → $6.5
Tensor Art Pro → $10
NovelAI Opus → $25

That’s around $45 per month.

Tools
  • NovelAI → image generation (paid site, my usual setup is v4.5, 28 steps, Guidance 5.8, Euler Ancestral)
  • Tensor Art → video genearation (paid site, my usual setup is WAN2.2-i2v-a14b-low-noise, CFG Scale 1, 720p, 16fps. When trying prompt 8 steps, 3 seconds. Once good prompt 13 steps, 5 to 8 seconds)
  • RealESRGAN → AI images upscaler to 4K
  • Paint dot NET → Image editor (for adding text, signature and rescale back the image)
  • DaVinci Resolve → Video editor (for merging MP4s, adding transition and logo, speed control...)
  • Shutter Encoder → Video encoder (You can extract AND upscale in the same time every frames from a MP4. This useful for creating a HQ GIF, or you can even reimport all the upscaled frames into Shutter Encoder and make a MP4 from them. Choose AI animation x2/x4 in source for that. This allows you to have a free AI video upscaler software)
  • Gifski → GIF creator (If you plan to post your animated GIF as Pixiv Ugoira note that changing the --quality from Gifski to anything lower than 100 will make the GIF corrupted for Pixiv. So you have to play with other setting like width for keeping you GIF under 15 MB)
Most used tags

Soon! I'm currently listing and ordering them.

gerhard96

Edit

Pub: 12 Oct 2025 16:21 UTC

Edit: 24 Nov 2025 09:53 UTC

Views: 340