Did you know that there was no graphic design, from kinetics to y2k, in the LAION dataset of Stable Diffusion? Yet people are still finetuning portraits and hentai generation!This is the first SD model trained specifically for motion design and music video production. Here's an example:<img src="https://imagecache.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/25e9f8d5-c29f-4475-e6d0-f2893edac500/width=525/25e9f8d5-c29f-4475-e6d0-f2893edac500" />You can check this breakdown if you're interested to see more details:<div data-youtube-video><iframe width="640" height="480" allowfullscreen="true" autoplay="false" disablekbcontrols="false" enableiframeapi="false" endtime="0" ivloadpolicy="0" loop="false" modestbranding="false" origin playlist src="https://www.youtube.com/embed/l4hDyLOGNjM" start="0"></iframe></div>The model is very capable in high budget MV production, posters, album covers, etc. Still, I thought it would be right to publish all the materials in public domain, as a tribute to the open source community of Stable DiffusionFeel free to clarify the workflow stuff:)Aerial v1.0<ul><li>Specialized on Deforum and img2img stylization</li><li>Trained on graphic design and art by me</li><li>Free to use, no licensing or credits required, forever</li></ul>Use the word "aerialstyle" to specifically trigger the style of the modelI don't recommend using the model in txt2img because of the abstract nature of the model. nonetheless, below are examples of images generated in txt2img. To get really interesting results though, you can generate a base image with the other model and give the image a style with Aerial, or just use it with ControlNet. This model is really good with CNs:)

Files with SHA256: 115f259377bd37c0a4b47e01d7e2b125fea56111826c1a2ba8e3ae2e4c61a5f3