Verified: 5 months ago
SafeTensor
Type | |
Stats | 230 151 |
Reviews | (25) |
Published | Jul 24, 2024 |
Base Model | |
Trigger Words | FILM PHOTO 90S FLASH PHOTO |
Hash | AutoV2 1D1935EDB2 |
推荐与LowRa或epiNoiseoffset结合使用,LowRa:0.4~0.7,epiNoiseoffset:0.5~1.5。使用add_detail:0.25左右会让图片更加高清。
LowRa: 0.4 to 0.7, epiNoiseoffset: 0.5 to 1.5, and epiNoiseoffset: 0.4 to 1.5 are recommended. Using add detail:0.25 or so will make the image more HD.
请不要使用Renhang作为提示词,我本来把这个词作为触发词,但今天检查到这个词在模型中存在,触发词被污染了。
Please do not use Renhang as a prompt word, I had this word as a trigger word but checked today that it exists in the model and the prompt is contaminated.
文生图更加适合室外场景,图生图和controlnet效果更好
Wenshengtu is more suitable for outdoor scenes, and graphs and controlnet work better
使用任航的750张摄影作品作为训练集,出片暂时不是很稳定,但有些图片效果非常写实
Using Ren Hang's 750 photographs as a training set, the output is not very stable for the time being, but some pictures are very realistic.