Pokemon
Obsession Series
There’s an allure to a man who embraces his love for Pokemon, and not afraid to express his genuine interests — even if it seemed silly and childish. It exposes an endearing vulnerability, and invites others to be authentic and playful. This “big boy” energy, a childlike wonder, is something very endearing to me — in ways that I can’t quite put into words — but is mainly why I have created this series of images.
Pokemon, Embracing the Inner Child. Rendered with Stable Diffusion, Airfuck’s Brute Mix, and Control Nets with SDXL images.
The Obsession series has a backstory stretching all the way to May 2023, with Pokemon as the OG. I never realized the series with Pokemon, mostly out of fear that most won’t relate to it. I posted that original Pokemon image that inspired the series in July, and the reactions made me realize that my assumption was wrong, so I turned it into a series today.
Background
I’d like to explain where the Mickey Mouse came from — it all started with Pokemon. A long time ago, I was browsing an Instagram tag, and found an image of a guy lying in bed with his Pokemon stuffed toys. I don’t remember what his name was anymore — it was a long time ago. All I remembered was that he was a fashion designer, and he posted his photo with Pokemons with all the regular hashtags that he typically used, even if the subject matter didn’t directly relate to the hashtag.
Before the Instagram user replied to the comment about the image, I didn’t know that the hashtag association was not intentional. I left him a comment along the lines of “Pokemon as a fetish? What a concept!” Only to find out later that he was simply reusing old tags without bothering to change them. But that unexpected association did put me into a d deep analysis mode, trying to think of ways how Pokemons could be associated with the concept of fetish.
After a brief think, my hypothesis / theory is that the word fetish can simply be interpreted as a form of obsession — that fetishism of any objects is no different than obsession with specific objects and subject matters, and potentially the only difference is that obsessions are more widely accepted, whereas fetishism tends to be more niche, and might not be as generally tolerated by all.
But let’s say that Pokemon somehow represents something that’s not generally popular culturally, then really, that passion you have for Pokemons is really not that much different than fetishism, especially if it produces a state of physical reaction.
With that concept in mind, I made a bunch of renders inside Midjourney, depicting a man who is obsessed with Pokemon, and how that life must be like for him — surrounding himself with Pokemons in his bed, and sleeping with them at night.
Images
From Pokemon to Teddy Bears
I made a few renders but thought that Pokemon might be too niche for the gay audience, so subsequently I realized this concept as a guy who’s obsessed with bears instead.
From there, I created three series back in May:
Instagram Story
Fast forward to July, a friend suggested that I start posting one-shots outside of the regular series, so I posted that Pokemon image to an Instagram story. To my surprise, it was very well received, and that got me thinking that I really should make a complete series out of it — after all, that is actually where the idea of obsession came from.
SDXL
My early experiments with SDXL has been very promising, and I’m very impressed with its dreamability, so I thought that I would remake the Pokemon series using SDXL. I made a bunch of them but ultimately decided not to post them — while the renders were nice, they lack the emotional drive that I sought. I made a ton of them — 93 to be exact — but if it’s not successful then I would rather not have them affect the perception of my work.
Airfuck’s Brute Mix
Fast forward to today, and Airfuck just released a new checkpoint — Airfuck’s Brute Mix — what better way to test out this the model than to work on a series with it — so that’s why I used it to realize Pokemon again.
Even though I really like the render quality of the model, an issue with models based in the SD 1.5 base model is that they really lack dreamability in composition — so what I did instead is use the SDXL renders I had to influence the color and composition with these new renders. How? Control Net, of course!
And that’s how this new series was born — it’s less about obsession, but how I found the charm in a man who’s not afraid of embracing his Inner Child. There’s something very sexy about a man who is not afraid to show his cute side. It’s the “big boy” impression that I get when men in their 30s and 40s showing childlike wonders. To me, that’s a really wonderful quality — because it demonstrates that the men are in touch with their emotions, and not take everything so seriously. Perfect husband material if you ask me.
Technical Details
Technically, I used a mix of Control Nets. There are five total, and I used a variation of them depending on the source SDXL image that I picked to drive the composition and color.
- Control Net Depth
- Control Net Shuffle
- Control Net Soft Edge
- Control Net Reference Only
- Control Net Open Pose
For all of these, I applied:
- Weight: 0.5
- Starting Step: 0
- Ending Step: 0.5
- Resolution: 1024
Eventually, what I settled with were these 3. There is a very strong reasons for this but it would make this post extremely long given the amount of images included here. I will write about this in a separate post when I have time.
Most of these were applied so that there will be variations from the SDXL source images. You may wish to check out my article about Control Net parameters to understand why I picked these values. Talking about these values in detail would be fairly involved. Since this post is already very long, if I have time, I will explain them in detail in a different post in the future.
Technical Parameters for the First Image
Because there are so many variations for these, the parameters are for the first image in the set:
Text prompts in Stable Diffusion. See Detailed Technical Parameters for the full settings.
- Steps: 30
- Sampler: DPM++ 2M Karras
- CFG scale: 5
- Size: 512x512
- Model hash: 8ff8347bb0
- Model: airfucksBruteMix_v10
- Denoising strength: 0
- CN 1: Shuffle, Weight 0.5, Start 0, End 0.5, Res 1024, Balanced
- CN 2: SoftEdge, Weight 0.5, Start 0, End 0.5, Res 1024, Prompt-priority
- CN 3: OpenPose, Weight 0.5, Start 0, End 0.5, Res 1024, Balanced
- Hires upscale: 1.5, 768x768
- Hires steps: 10
- Hires upscaler: 4x_NMKD-Siax_200k
- face_editor_enabled: True
- A1111 Version: v1.5.1
- Post: Topaz Gigapixel HQ 4x, 3072x3072
Images
Image Index
Stable Diffusion with Airfuck’s Brute Mix. Series Images.
Stable Diffusion with SDXL. Source SDXL Images for Control Nets.
Midjourney v5. OG Pokemon renders in May.
Stable Diffusion: Failed Renders
These images are fails to be due to various reasons — I plan on covering them in a technical article in the future. Some of them are quite fun — mostly due to how Control Nets behave. But to understand each one, you’ll need to understand how Control Net work, and that’s a very deep topic to cover. So I’ll cover them in the future. For now, just have fun looking at them.
Detailed Technical Parameters
Because there are so many variations for these, the parameters are for the first image in the set:
- Steps: 30
- Sampler: DPM++ 2M Karras
- CFG scale: 5
- Seed: 1240913996
- Size: 768x768
- Model hash: 8ff8347bb0
- Model: airfucksBruteMix_v10
- Denoising strength: 0.4
- ControlNet 1:
- preprocessor: shuffle
- model: control_v11e_sd15_shuffle 526bfdae
- weight: 0.5
- starting/ending: (0, 0.5)
- resize mode: Crop and Resize
- pixel perfect: False
- control mode: Balanced
- preprocessor params: (1024, -1, -1)
- ControlNet 2:
- preprocessor: softedge_pidinet
- model: control_v11p_sd15_softedge a8575a2a
- weight: 0.5
- starting/ending: (0, 0.5)
- resize mode: Crop and Resize
- pixel perfect: False
- control mode: My prompt is more important
- preprocessor params: (1024, -1, -1)
- ControlNet 4:
- preprocessor: openpose
- model: control_v11p_sd15_openpose cab727d4
- weight: 0.5
- starting/ending: (0, 0.5)
- resize mode: Crop and Resize
- pixel perfect: False
- control mode: Balanced
- preprocessor params: (1024, -1, -1)
- Hires negative prompt: "..."
- Hires upscale: 1.5
- Hires steps: 10
- Hires upscaler: 4x_NMKD-Siax_200k
- TI hashes:
- Asian-Less-Neg: 22d2f003e76f
- bad-hands-5: aa7651be154c
- by boring_e621: 8b580f3de3e1
- easynegative: c74b4e810b03
- face_editor_enabled: True
- face_editor_face_margin: 1.6
- face_editor_confidence: 0.97
- face_editor_strength1: 0.4
- face_editor_strength2: 0
- face_editor_max_face_count: 20
- face_editor_mask_size: 0
- face_editor_mask_blur: 12
- face_editor_prompt_for_face: ""
- face_editor_apply_inside_mask_only: True
- face_editor_apply_scripts_to_faces: False
- face_editor_face_size: 512
- face_editor_use_minimal_area: False
- face_editor_ignore_larger_faces: True
- face_editor_affected_areas: F;a;c;e;Face
- face_editor_upscaler: None
- face_editor_tilt_adjustment_threshold: 40
- Version: v1.5.1