这期介绍一个非常快的模型!主打一个高效!大家都知道FLUX很吃显存, 但是最近一个模型引起大家热议!它就是shuttle-3-diffusion。
shuttle-3-diffusion介绍
shuttle-3-diffusion由一家名为ShuttleAI的公司基于Flux.1 可商用模型Schnell开发的模型,它个步骤从文本提示创建详细和多样化的图像。它在图像质量、排版、理解复杂提示和资源效率方面提供了增强的性能。
它有很多个版本,模型在训练期间部分去蒸馏。当使用超过10个步骤时,它进入“细化模式”,在不改变构图的情况下增强图像细节。
这家公司训练了将近13个模型。包括bfloat16、GGUF、fp8多个变体版本。
下载地址
这是它的镜像站地址:https://hf-mirror.com/shuttleai
对应的模型大家点击即可进入它的下载页面。
推荐大家使用它GGUF和FP8,
GGUF版本:https://hf-mirror.com/shuttleai/shuttle-3-diffusion-GGUF
FP8版本:https://hf-mirror.com/shuttleai/shuttle-3-diffusion-fp8
下载后,放置到unet下即可:
测试效果
首先我们先测试下4步出图,搭建一个文生图工作流:
提示词输入:
A photo-realistic shoot from a front camera angle about a woman in a long, flowing, light green dress surrounded by large, green leaves and white lotus flowers. the image also shows a young woman standing in the middle of the image, wearing a light green, sleeveless, floor-length gown with a slit, adorned with delicate white flowers. she has a slender physique, fair skin, and a serene expression. her hair is styled in loose, dark locks, and she is wearing white sandals. she is standing in front of a backdrop of large, leafy plants, with a plain, light beige wall behind her. the woman is facing the viewer, looking directly at them with a neutral expression, and her eyes are framed by her closed mouth. her hands are clasped together in a symmetrical manner, with one hand resting on her chest and the other holding a small, delicate flower. the overall aesthetic is elegant and minimalistic, with the woman's dress and lotus flower adding a touch of nature to the scene.
同样我们测试10步出图:增加细节
提示词:
A photo-realistic shoot from the side about a young woman sitting on a chair holding a rose, wearing a red ribbed tank top and a blue denim skirt. the image also shows a table with a vase of red flowers and a vintage fan in the background. on the middle of the image, a 20 years old asian woman with light skin, black hair styled in long wavy hair, and a serious expression is sitting on the chair. she is holding a red rose in her right hand and wearing large pink earrings. the woman appears to be looking directly at the camera with a neutral expression. the background features a green wall with a black and white photograph of a man. the lighting is soft and warm, creating a cozy atmosphere.
4步结果:
10步结果:
卡通风:
提示词
A digital illustration shoot from a front camera angle about a young woman standing in a lush garden with two small dogs and a large yellow sun in the background. the image also shows vibrant colors and a whimsical atmosphere. on the middle of the image, a 20-year-old asian woman with light skin, long brown hair, and brown eyes appears to be holding a bottle of perfume in her hand. she is wearing an orange sweater, a black skirt, black shoes, and a black hat with a red ribbon. her expression is neutral, and she is looking directly at the viewer. she has a slim body and is wearing black socks and black shoes. her hair is styled in a long hair style with bangs. her eye color is not specified, but her eye direction is looking at viewer. her body is positioned in the center of the frame, with her feet out of frame. on either side of her, two small brown and white chihuahuas are standing. the background features lush greenery, red flowers, and tall trees, with a large, glowing yellow sun behind her and a few butterflies fluttering around.
试下加入pulid flux看支不支持。
我们再来一次。同样用寡姐当我们的模特。
可行!不错,直接提升效率。
由于我锁定了种子,所以它基本只换了脸,我们可以再试一次。
输入提示词:
A vibrant and colorful digital illustration shoot from a portrait camera angle about a young woman with short, black hair styled in a bob cut, wearing sunglasses and holding a green iphone, surrounded by a green background filled with various fruit and vegetable-themed illustrations. the image also shows playful and whimsical elements such as fruits, leaves, and smiley faces. on the middle of the image, a 1woman, who appears to be a young adult, with a fair complexion, green eyes, and a small nose, is standing in the middle, looking directly at the viewer with a neutral expression. she is wearing a black shirt with a floral pattern and a choker necklace, and has a pair of earrings on her ears. her hair is styled in short, wavy black hair, and she is holding a bright green iPhone in her right hand, with her left hand resting on her collarbone. the overall style is reminiscent of japanese anime and manga, with bright colors and playful elements.
好的。这次的模型测试先到这里,后面我们可以基于这个模型,使用我们三阶采样工作流来提升图片画质,以及画人像的质量。
这里直接将该软件分享出来给大家吧~
1.stable diffusion安装包
随着技术的迭代,目前 Stable Diffusion 已经能够生成非常艺术化的图片了,完全有赶超人类的架势,已经有不少工作被这类服务替代,比如制作一个 logo 图片,画一张虚拟老婆照片,画质堪比相机。
最新 Stable Diffusion 除了有win多个版本,就算说底端的显卡也能玩了哦!此外还带来了Mac版本,仅支持macOS 12.3或更高版本。
2.stable diffusion视频合集
我们在学习的时候,往往书籍源码难以理解,阅读困难,这时候视频教程教程是就很适合了,生动形象加上案例实战,一步步带你入坑stable diffusion,科学有趣才能更方便的学习下去。
3.stable diffusion模型下载
stable diffusion往往一开始使用时图片等无法达到理想的生成效果,这时则需要通过使用大量训练数据,调整模型的超参数(如学习率、训练轮数、模型大小等),可以使得模型更好地适应数据集,并生成更加真实、准确、高质量的图像。
4.stable diffusion提示词
提示词是构建由文本到图像模型解释和理解的单词的过程。可以把它理解为你告诉 AI 模型要画什么而需要说的语言,整个SD学习过程中都离不开这本提示词手册。
5.SD从0到落地实战演练
如果你能在15天内完成所有的任务,那你堪称天才。然而,如果你能完成 60-70% 的内容,你就已经开始具备成为一名SD大神的正确特征了。
这份完整版的stable diffusion资料我已经打包好,需要的点击下方插件,即可前往免费领取!