For Z-Image: Not particularly happy with how this turned out but I thought it was fine to release. Probably won't get better results until the base version is released.
Hard Edge style pixel art. Use "pixel art" or "pixel art style" as the trigger word.
Should scale down perfectly by downscaling 4X with nearest neighbor (not as perfect for Z-Image version)
Check out my other model, Soft Pixel Art, if you prefer a softer, more detailed style:
Soft Pixel Art - v1.0 | Stable Diffusion LoRA | Civitai
Tip: Combine both of these LoRAs at 0.5 and 0.5 to get an in-between look!
Description
FAQ
Comments (4)
hello, could you please share your training details? I am wondering how to train a excellent lora like yours, I use kohya gui and Fluxtrainer(comfyui node) to train, but I can't get the good result like you
SECONDED!! I have been training pixel Flux loras but this one remains the cleanest I've found
https://lorainfo.tools/ you can drag his loras on to this tool to find out more!
@kromdario Thanks for your help, but I dray this pixel lora to this tool, it shows the base model is sd1.5?
tool result:
{
"sshs_model_hash": "3A4511A9304A8D6440F349BDDEF0FFDA7A1CEC82618821830159B0C3E61ACA3C",
"name": "test_lora",
"ss_output_name": "test_lora",
"format": "pt",
"sshs_legacy_hash": "e3b0c442",
"training_info": "{\"step\": 4000, \"epoch\": 153}",
"version": "1.0",
"ss_base_model_version": "sd_1.5",
"software": "{\"name\": \"ai-toolkit\", \"repo\": \"https://github.com/ostris/ai-toolkit\", \"version\": \"0.1.0\"}"
}
Details
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.