
(SDXL Lora) - V3 - Pixelized Art - Intermediate Edition
Adds a "Pixelized" (and color depth reduction) effect as seen in blocky, early 90's /Gameboy Advance artwork.
This third version is a completely retrained version based on the experience I obtained while training the V1 and V2 versions, granting a high resolution output image with very flexible, and high pixelation possible.
Also expect a more "edge" pronounced pixel effect (with reduced color depth).
Personally I find this version to give the best pixelated results compared to V1 and V2.
(See example images)
-
Use "p1x3l" as a prompt word.
-
Flexible pixelization by adjusting Lora Strength ( 0.7 < Lora Strength < 1.4 )
-
Use 1.3 as base pixelization
-
-
SGM + Uniform and DPMPP_2S_Ancestral + Karras work well
-
CFG 3 /20 Steps /No refiner
-
See examples pictures.
-
dreamshaperXL seems to be a good checkpoint
-
See examples pictures and comparison pictures below
-
First version (V1): https://civitai.com/models/527637/sdxl-lora-pixelized-art-lora-style
-
Second version (V2): https://civitai.com/models/536161/sdxl-lora-v2-hd-pixelized-art
SD1.5 version available here:
https://civitai.com/models/550585/sd-15-lora-v3-pixelized-art-intermediate-edition
Workflow to convert existing pictures with Lora(s) (to pixelized for example) here:
https://civitai.com/models/539896/sd-15-workflow-apply-lora-effect-to-image
描述:
训练词语: p1x3l
名称: PixelV3-xl1024-v7a-0.safetensors
大小 (KB): 223097
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success