向那些令人印象深刻的画风致敬，他们开创了一种艺术风格，引领一部分人以奇特或者神奇的角度去观赏这个世界。
我们的世界从不缺失美，只是缺少一双发现美的眼睛。
Pay tribute to those impressive painting styles, they created or promoted an artistic style, leading some people to view the world from a strange or magical perspective.
Our world never lacks beauty, but lacks a pair of eyes to discover beauty.
Description
2023年4月23日，是《饥荒》这个生存游戏的十周年纪念日。
《饥荒》是一款由加拿大公司科雷娱乐开发的开放世界生存游戏。
这个游戏围绕在一位名叫威尔逊的科学家，他发现自己身处黑暗阴暗的世界，必须尽可能长时间生存。为此，玩家必须保持威尔逊的健康、食物和精神稳定来避免了各种超现实和超自然的敌人杀死和吞噬他。
随着十年间的不断发展，科雷将生存世界以奇妙的线条感和粗糙而精准的手绘风格展现给大家，营造了一个畅玩十年依旧经典的《饥荒》世界。
April 23, 2023 marks the tenth anniversary of the survival game, 《Don't Starve》.
《Don't Starve》 is an open world survival game developed by the Canadian company Klei.
This game revolves around a scientist named Wilson who finds himself in a dark and gloomy world and must survive for as long as possible. To do so, the player must keep Wilson healthy, fed and mentally stable to avoid the various surreal and supernatural enemies that kill and consume him.
As the decade progressed, Klei brought the world of Survival to life with a wonderful sense of line and a rough and precise hand-drawn style, creating a world named as 《Don't Starve》that is still classic after ten years of play.
FAQ
Comments (8)
Too bad there are no prompt examples
very simple to use, just 1boy, <lora>, and then genearate it. All my lora pursuits a specific goal, that is, it will be useful when loaded.
dude, it's not a lora, it's a lyco. Move it from the Lora folder to the Lycoris folder, then try this "(best quality:1.1) 1girl, a young girl, cute face, holding an umbrella, dancing in the rain, <lyco:DontStarve饥荒:1.0>" I just used Easynegative and revAnimated model. It's a neat effect.
my bad. guess the Google translation doesnt work properly. the author said it is lyconris. somehow tried it as lora. that is why it didnt work. my apology. :-(
still have a favour to ask. pls provide the prompt for other models in this post ie wildermyth. thanks in a bunch.
do you mean hypoxia?
What you said makes sense. If my sample database were large, or when making a model, high dim training is a good method to reduce underfitting. But when my samples are not many enough, overfitting is an important factor I have to consider. The occurrence of Curse of Dimensionality will also make the LoRA uncontrollable, which is also something worth noting. No matter whether it is 128dim or 32dim, the method of dimensionality reduction and convolution can reduce the factors of measuring distance. Therefore, after considering my hardware equipment and database, 32dim is a good choice for me. And, of course, 16dim and 8dim alpha do have XOR gate effect. and this effection would be more in SD2.X model training. But relax, this is a lora.
Tried this on 10 different models, could not get it work. <shrug>. Too bad, looks like neat output.
EDIT: Bro, this is NOT a basic Lora! It's a LyCORIS, no wonder nobody could get this to work. You have to change the category. Also, for some reason when I posted my pictures it did not pick up the generation data from my png? Maybe because of the chinese characters?
Using it as a LORA or LyCORIS gave exactly the same results for me.