Type | |
Stats | 290 6,937 4 |
Reviews | (20) |
Published | Jan 10, 2024 |
Base Model | |
Training | Steps: 4,440 Epochs: 10 |
Usage Tips | Clip Skip: 2 Strength: 1 |
Hash | AutoV2 980B06E97D |
This LoRA learned on low resolution suitable use for about little upper to double resolution. Trained resolution 128 and 256 for generation of HQ LowRes Images.
Trigger words unnecessary
class word: 256, alice in wonderland, angelic, hands, realistic, unrealistic, doll, underwater
Marge to models (for version (above)v2.0, now testing)
Before in the first designed merges: near max strength merging to basemodel as much as possible. New model could drawn for low resolutions. However, basemodel is style changed some little.
(eg: merge_lora.py --sd_model XXX --save_to XXX --models 256-lycoris --ratios 0.5)
V1.0 Realistic (Testing)
If merge of high resolution, need to merge 512(to less then 1024) resolution first as an intermediate balancer when as 256 LoRA have strength(equality merges neccesary??? poor drawn eye with full body if bad mearges).
highPixelar: https://civitai.com/models/253298
🌟 How to get true detailed definition feeling on SD1.5 low resolution?(nearly 2.1768 & SDXL feeling)