可以出各种像素小人人儿!可以配合其他lora来玩!也可以出像素场景!权重0.5就行!
画人像素推荐512X768,高清修复1.5.重绘幅度0.4,再高就没有像素感了。
画建筑推荐768X768,高清修复1.2
触发关键词:pixel

You can make all kinds of pixel charactor! You can play with other lora! You can also create pixel scenes! A weight of 0.5 <local lora name:0.5>will do!
512X768 pixel is recommended for drawing character, hires fix: Upscale by 1.5, Denoising strength 0.4, no pixel feeling if it is higher.
Drawing buildings recommended 768X768, hires fix Upscale by1.2,
Trigger Words: pixel
いろんなピクセルキャラが作れます! 他のローラと遊べる! ピクセルシーンも作れます! 0.5 の重みで十分です。また、
人物を描くなら512X768ピクセル推奨、HDリペア1.5、再描画範囲0.4、それ以上だとピクセル感なし。
建物の描画推奨 768X768、HD リペア 1.2
-- -- -- -- -- -- --
3.0整合了前两版的优点,增加大模型兼容性的同时,现在能生成更加精细,符合现代审美及设计感的像素画了。同时场景生成方面也获得了强化。建筑,室内,SLG等等都可以尝试下。
像素画起源于上世纪末,是电脑/游戏机/显示器机能不足下妥协的产物。力求在有限的算力和分辨率限制下绘制出精细的画作。21世纪后,渐渐的硬件水平提高了,像素画也获得了精细度的解放。同时3D技术大行其道,像素艺术家们只能在独立游戏和任天堂掌机上继续创作,同时像素画也作为一种风格继承了下来,由于细节的缺失,反而给人一定的想象的空间。以歧路旅人为代表的HD2D则是其现代精神继任者。
以此来缅怀那个像素艺术的黄金年代。
This version integrates the advantages of the previous two versions. While increasing the compatibility of large models, it can now generate more detailed pixel paintings that meet modern aesthetics and design. At the same time, scene generation has also been enhanced. Architecture, interior, SLG, etc. can all be tried.
Pixel art originated at the end of the last century, and it is the product of compromise under the insufficient performance of computers/game consoles/monitors. Strive to draw fine paintings under the limitation of limited computing power and resolution. After the 21st century, the level of hardware has gradually improved, and pixel art has also been liberated in fineness. At the same time, 3D technology is popular, and pixel artists can only continue to create on independent games and Nintendo handhelds. At the same time, pixel art has also been inherited as a style. Due to the lack of details, it gives people a certain space for imagination. by HD2D represented by Octopath Traveler is its modern spiritual successor. In order to remember the golden age of pixel art.
このバージョンは、前の 2 つのバージョンの利点を統合し、大規模なモデルの互換性を高めながら、現代の美学とデザインを満たす、より詳細なピクセル ペインティングを生成できるようになりました。 同時に、シーン生成も強化されました。 建築、インテリア、SLGなど、すべてお試しいただけます。
ドット絵は前世紀末に誕生したもので、パソコン・ゲーム機・モニターの性能不足による妥協の産物です。 限られた計算能力と解像度の中で、精巧な絵を描くように努めます。 21世紀以降、ハードウェアのレベルが徐々に向上し、ドット絵も細かさが解放されてきました。 同時に、3D 技術が普及し、ピクセル アーティストは、独立したゲームや任天堂のハンドヘルドでのみ作成を続けることができます. 同時に、ピクセル アートもスタイルとして継承されています. 詳細が不足しているため、人々に想像力のための特定のスペース。 に オクトパス トラベラーに代表される HD2D は、その現代的なスピリチュアルな後継者です。 ピクセルアートの黄金時代を思い出すために。
-- -- -- -- -- -- --
2.0降低了拟合度。能更好的适配各种大模型。更能出结合场景的角色,和更丰富的画面。相对的像素感没那么严格了,比如像素描边少了,更色块化,所以说是像素风更合适,不再是像素画了。喜欢传统像素的推荐继续用1.0.
2.0 reduces the fit. It can better adapt to various large models. More characters that can combine scenes, and richer pictures. Relatively, the pixel sense is not so strict, for example, there are fewer pixel strokes and more color blocks, so it is more appropriate to say that the pixel style is no longer a pixel painting.Those who like traditional pixels are recommended to continue to use 1.0.
2.0 ではフィット感が低下します。 さまざまな大型モデルによりよく適応できます。 シーンを組み合わせることができるより多くのキャラクター、およびより豊かな写真。 比較的、ピクセルセンスはそれほど厳密ではありません。たとえば、ピクセルストロークが少なく、カラーブロックが多いため、ピクセルスタイルはもはやピクセルペインティングではないと言う方が適切です.従来のピクセルが好きな方は、引き続き 1.0 を使用することをお勧めします。
Description
这一版整合了前两版的优点,现在能生成更加精细,符合现代审美的像素画了。
FAQ
Comments (46)
场景生成很不错,唯一问题就是不怎么像素,像素的风格有了,可是“分辨率”不够低,没有像素格子
还没试场景,不过角色都还蛮不错的,就是不太知道怎么调整像素的大小,谢谢分享
Very useful pixel art, but still a shortcoming of AI production, it looks ok, but the details are poor, but it is a good example of painting, well done!
безусловно необычная и очень красивая лора, но йоу она просто супер непослушная, и не хочет делать интересные фоны
сделай послушную
How do I use HD repair 1.5??
its hires fix
RuntimeError: The model currently loading is not a valid Stable Diffusion checkpoint.
+1
THIS IS NOT A STABLE DIFFUSION CHECKPOINT MODEL, it's a LoRa model, and works perfectly
yes,it is not ckpt,is a lora,
Should I use any VAE? The color of my pictures seems not so light.
yes try different vae,
try pastel diffusion vae, i had some pretty neat colors
@Fragi1e Can you tell me what VAE you use?
大佬能求个联系方式吗,这边小团队创业做AI画图,不知道有没有兴趣加入我们一起,有兴趣的话可以发我邮件 [email protected] 详聊
RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cpu and cuda:0! (when checking argument for argument index in method wrapper_CUDA__index_select)
我无法使用这个模型,o(╥﹏╥)o,
我不知道为什么,但是我真的很希望能玩一玩像素风.有没有人帮帮我?我尝试在谷歌中搜索,有人也跟我遇到同样的问题,但是我始终找不到我想要的答案.
I don't know why, but I really want to play with pixel wind. Can anyone help me? I tried to search in Google, and someone had the same problem as me, but I couldn't find it. answer you want.
looks like u need use only gpu to run.try different version of my lora(i upload 3ver),
I have used more than a dozen models without any problems, but when I use the Stable Diffusion model (ckpt), which is the option in the upper left corner, to load or submit your pixel model, it will flash for a while, and then there will be no change.
According to your reminder, I tried the 3 pixel models you posted, but none of them worked.
There is such a prompt in the console:
RuntimeError: The model currently loading is a LoRA instead of a Stable Diffusion checkpoint.
__________________________________________________________
Loading weights [a4a3da4121] from P:\novelai-webui-aki-v3A\models\Stable-diffusion\pixel_5.safetensors
changing setting sd_model_checkpoint to pixel_5.safetensors [a4a3da4121]: RuntimeError
Traceback (most recent call last):
File "P:\novelai-webui-aki-v3A\modules\shared.py", line 509, in set
self.data_labels[key].onchange()
File "P:\novelai-webui-aki-v3A\modules\call_queue.py", line 15, in f
res = func(*args, **kwargs)
File "P:\novelai-webui-aki-v3A\webui.py", line 205, in <lambda>
shared.opts.onchange("sd_model_checkpoint", wrap_queued_call(lambda: modules.sd_models.reload_model_weights()))
File "P:\novelai-webui-aki-v3A\modules\sd_models.py", line 505, in reload_model_weights
state_dict = get_checkpoint_state_dict(checkpoint_info, timer)
File "P:\novelai-webui-aki-v3A\.launcher\pyinterop.hfkx1kkk0g4q7.zip\sitecustomize.py", line 236, in wrapper
raise exception_wrap(RuntimeError("The model currently loading is a LoRA instead of a Stable Diffusion checkpoint."))
File "P:\novelai-webui-aki-v3A\modules\sd_models.py", line 505, in reload_model_weights
state_dict = get_checkpoint_state_dict(checkpoint_info, timer)
____________________________________________________
I can load other people's models, but yours can't....... XD
我知道为啥了兄弟,你把Lora当大模型了把。lora不在左上角那选。在生成按钮底下那块一个红色的小图标。。。你放的位置也不对,那个位置是放大模型的,lora放隔壁的lora文件夹。。。基本144M以下的都是小模型不能直接用的,,看看教程把
十分感谢你能回答我的问题,我终于处理好了,前几天没显卡没空试,我估计是因为我放错文件夹了,之前我也尝试放入lora的文件夹,哈哈哈哈.想问下配合什么大模型会好一些?
How to generate sequence frame animation?
I wish there would be a version 4 that doesn't bias the scenery as much towards medieval elements. With better support for things like modern and future scenes.
My lora have stopped working. yesterday they were working fine now not anymore. what can i do?
How to avoid text on the result image? Already added the negative prompt: text.
How can I get this to not show shadows on the feet? "shadow" in the negative prompt does not achieve it
What base model did you use for this? or other loras? with default sd 1.5 i get just pixelated real life haha
AnyLora model
As cool as it is,I actually found a different use for it, at 0.1-0.2 strength it won't do anything visually,but it still frames the character in the center of the image,something I had a problem of doing without OpenPose controlnet.
我跟着你说的这样子生成的图片有一些多颜色了,可能是大模型的问题,作者你用的是什么SD模型,可以分享出来吗?
图片右下角的感叹号,点一下,数据,使用模型,关键词,都有
Perfect results in A1111, but it doesn't work in ComfyUI at all, i dont know why. As a result, we have ComfyUI, which provides perfect animation, and A1111, which provides perfect results, but works so bad with animations. Sad story :(
這個,可以做行走十二宮格素材嗎?RPG Maker用的。
我很喜欢你的工作,太棒了。这是我第一个评论,送给你
为什么我的出图很模糊呢
用这个lora一定要开高清修复,不然正常出图完全没有像素的感觉,高清修复一下反而起效了变像素了,研究了好久才发现...
训练集里男女有不均衡吗,我提示词写1man出来的都是女生\(“ ̄□ ̄)/
Details
Files
pixel_f2.safetensors
Mirrors
pixel_f2.safetensors
pixel_f2.safetensors
pixel_f2.safetensors
pixel_f2.safetensors
pixel_f2.safetensors
mPixel_v30MordenPixel.safetensors
pixel_f2.safetensors
pixel_f2.safetensors
mPixel_v30MordenPixel.safetensors
pixel_f2.safetensors
pixel_f2.safetensors
pixel_f2.safetensors
a_pixel.safetensors
pixel_f2.safetensors
pixel_f2.safetensors
pixel_f2.safetensors
pixel_f2.safetensors
pixel_f2.safetensors
pixel_f2.safetensors
pixel_f2.safetensors
pixel_f2.safetensors
pixel_f2.safetensors
pixel_f2.safetensors
pixel_f2.safetensors
pixel_f2.safetensors
pixel_f2.safetensors
pixel_f2.safetensors
pixel_f2.safetensors
pixel_f2.safetensors
pixel_f2.safetensors
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.
















