Playstation Filter Using Stable Diffusion
The idea was to jump the trend of creating oldschool playstation styled photos
Already had a ridiculous profile picture with me with big sneakers standing on a pink background, the prime candidate to amp it up further.
Original prompt used midjourney (link to og prompt)
But could i run it myself? Good way to utilize the g9 server in the homelab as the 3070 on the gaming computer may not have enough vram for larger models and why stress the gaming machine if i could use the server with 256gb ram in the background
Link to webui docker
Automatic1111 and comfy UIs
Different models, parameters
Using sd1.5
First aha: CFG
Got background and ground: man big shoes light clothes sunglasses ps1 playstation psx gamecube game radioactive dreams screencapture 3d pixelated Steps: 80, Sampler: LMS, CFG scale: 16, Seed: 2788379268, Size: 512x512, Model hash: c6bbc15e32, Model: sd-v1-5-inpainting, Denoising strength: 0.75, Conditional mask weight: 1.0, Version: v1.7.0 00002-541820203.png
Switch to sdxl
Second aha: denoise strength man big shoes light clothes hands crossed
First decent result
Third aha: controlnet
man big shoes light clothes hands crossed
Img2img works best in tandem with the img also in a controlnet unit in my experience so that it preserves more of the original.
Cool lora, thanks to the creator LuisaP!
Steps: 24, Sampler: DPM++ 2M Karras, CFG scale: 8, Size: 512x512, Model hash: d8fd60692a, Model: sdxl-helloworld-xl, Denoising strength: 0.3, Lora hashes: "sdxl-psx: 97857fb508e5", Version: v1.7.0
ControlNet Unit 0 model: diffusers_xl_canny_full, pixel perfect: true, upload independent control image: true (same as img2img), control weight 1, control mode: controlnet priority
realistic, photoreal, highly detailed, (rich cat wearing gold chains), nightclub, dark, neon, contrast Negative prompt: cartoon, unrealistic, anime Steps: 21, Sampler: DPM++ 2M Karras, CFG scale: 7, Seed: 810480089, Size: 512x512, Model hash: d8fd60692a, Model: sdxl-helloworld-xl, ControlNet 0: "Module: depth_midas, Model: diffusers_xl_depth_full [e3b0c442], Weight: 0.9, Resize Mode: Just Resize, Low Vram: False, Processor Res: 512, Guidance Start: 0, Guidance End: 1, Pixel Perfect: True, Control Mode: Balanced, Hr Option: Both, Save Detected Map: True", Version: v1.7.0 controlnet 0: depth
then separate face
Photoshop process gif with previosly mentioned pics
add noise, grain and pixellate, final result: