The WaiSnackMix Bounty and My Special Honorable Mention - Trigger for this post
In the WaiSnackMix Bounty I narrowly missed the podium.
Yet I received a truly wonderful special honorable mention from @MGHerder. He admitted that with better insight into my workflow, third place would have been well within reach. He even published a follow-up article today when he realised that the Sally Tsung versus ork fight (Shadowrun theme) was also one of mine, and some other stuff he didn't see yesterday.
It seems some of my creations resonate deeply with many users. What I still lack is the reach and visibility that others already possess. This article aims to bridge that gap and give you a genuine insight into my creative process and workflow.
My Manual and Flexible Workflow
As I mentioned in my 300-follower celebration post, I use ComfyUI only occasionally, mainly for checkpoints that do not play well with SD reForge. I rarely generate images directly on the platform, and when I do they are almost always SFW. Yellow buzz is hard to come by, and without tips I accumulate very little. The small amount I have goes toward NSFW video experiments or tipping other creators whose work genuinely impresses me.
Without ComfyUI, my workflow is entirely manual and follows no rigid sequence. That is worth noting.
Sources of Inspiration
I find a great deal of inspiration simply by scrolling through the post gallery. That is how I discover new styles that I then adapt and transform for my own purposes. The unfinished inksplash motifs, for example, originated exactly this way.
My primary sources of ideas, in no particular order, are science fiction and fantasy novels, films, pop culture, music, complex board games (my ongoing Gloomhaven and Frosthaven video series being a prime example), current events, and especially conversations with my children. Those conversations often lead to the most fascinating concepts.
NSFW as My Early Follower Magnet
Looking through my posts reveals a broad spectrum of themes. At the beginning, NSFW and erotic content served more or less as the kickstarter to attract my first followers. Without it, I would hardly have been noticed. Even then, I tried to create something more creative and interesting rather than yet another generic bimbo.
These days I derive far more pleasure from SFW work, but I still want to keep my NSFW audience engaged (pun very much intended). That is why I occasionally include a few bolder pieces.
Bounties and Idea Overload
Many of my best ideas emerge from participating in bounties. While I can only submit a limited number, my mind generates far more concepts than I can use. The surplus ideas need an outlet somewhere.
The “Make something sad” bounty, for instance, triggered a real creative explosion. It gave birth to the series of black-and-white images and videos featuring exactly one coloured element, which in turn inspired the inksplash series. That bounty was particularly challenging for many checkpoints, and I simply cannot resist such a test.
The most difficult bounties are those that allow only a single submission. Then the question remains: what do I do with the dozens of other ideas?
My Virtual Creative Partner: Grok
Whenever an idea takes shape, I turn to my new virtual friend, Uncle Musk or rather Grok, and ask what else I could do with it, preferably something with a slightly sinister edge.
I believe the way I collaborate with Grok deserves its own article someday. Although I find Musk and the X platform rather questionable, the tool itself is incredibly powerful. Why not use Gemini or Claude? The answer is simple: Grok is allowed to be spicy. The others censor far too aggressively. And I enjoy when even my SFW ideas carry a subtle erotic tension.
I feed Grok with a clear task, a description of my idea, and sometimes an image or style reference for analysis. We then discuss how to develop the concept further. Sometimes I simply ask for five different prompt variations. I test them quickly, realistic prompts with Grok/Imagine (Flux 1.D) and illustrative work locally with Pony Diffusion V6 XL or Illustrious XL checkpoints.
This process helps me separate the promising ideas from the rest. I have become a strong advocate of natural-language prompts. Modern models handle them exceptionally well.
Only when the core idea feels solid do I add LoRAs, sometimes purely because crossposting on CivitAI rewards me with additional visibility (sorry, but that is simply how the platform works).
Two LoRAs are nearly always present: the Lady Hands LoRA for realistic female hands, and either the Dramatic Lighting Slider or Chiaroscuro Lighting Style, because I love dramatic lighting, even in illustrative and anime pieces. With some checkpoints I also rely on the skin-colour slider, since many models are heavily biased toward Caucasian features.
And I should never forget to mention the lazy embeddings, which are a constant tool in almost every generation.
From Prompt to Generation
With the idea, concept and prompt ready, it is time to generate. I usually start with DPM++ 2M SDE or Euler A (preferably Karras or Exponential), using a random seed and creating four to eight images at once. Even on my RTX 4050 this only takes one to two minutes per image.
If the result is not quite right, I return to Grok, describe exactly what bothers me, and have him refine the prompt. The beauty of natural language is that the differences between old and new prompts become clearly visible. Small changes can have a powerful effect.
Refinement Phase
Once the composition feels correct, I move into refinement. I often switch checkpoints at this stage and use an intermediate result as input for ControlNet (mainly Canny or Depth; OpenPose has rarely delivered convincing results for me).
Sometimes I first make radical edits in paint.net, masking unwanted elements or correcting anatomy to save time on later inpainting.
These model switches are especially common with NSFW work. I often begin with an anime model because it offers more freedom in anatomy and posing, then finish the piece with a realistic model. The marble statue series, featuring sensual encounters between warm flesh and cold marble, for example, had to take exactly this route.
Final Steps and Video Creations
The final step is usually an img2img pass for upscaling. I rarely use Ultimate Upscaler. If I stayed with the same checkpoint throughout, a simple high-res fix with the final seed is often sufficient.
As you can see from my profile, I have recently discovered great joy in turning my creations into short videos using Grok/Imagine. That process, however, deserves its own dedicated article. Otherwise this one would become far too long.
I hope this glimpse into my workflow offers you some inspiration as well.
If it does, I would be grateful for a like or a bit of yellow buzz.
PS: I gave Grok this article to swallow and digest and let it create the prompt for the header image. You see how easy things can be.
PS 2: Today (21st of March '26) I published a more detailed explanation of how I utilize Grok as a swift tool in my workflow process. You can find it here.