Plug-ins will always be updated, including 90% of the popular plug-ins on the market. You will not miss useful plug-ins. Please read previous articles for previous plug-ins. If you don’t have time to follow the progress of SD updates, please follow me. Spend a few hours a month reading my articles and follow up on SD in the shortest possible time. It’s also my personal note
SD plug-in
sd-webui-animatediff
https://github.com/continue-revolution/sd-webui-animatediff
https://huggingface.co/guoyww/animatediff/tree/main
The main model is placed in the plug-in model path, lora is placed and used normally, and some prompt words are added.
inpaint-anything
https://github.com/Uminosachi/sd-webui-inpaint-anything
If there is no problem with the segmented object, click to get the mask, and then click below to send it to the picture; simply write some modified clothing prompts, such as red clothes, flowers, etc.
OK, the changeover is complete! ! !
sd-webui-EasyPhoto
https://github.com/aigc-apps/sd-webui-EasyPhoto
For training, just upload 10 photos. The parameters basically don’t need to be changed. The batch size can be adjusted upward according to your video memory, and the corresponding gradient accumulation should also be adjusted downward.
reasoning
The above is just a simple display. For details, you can refer to the website documentation, which is quite detailed.
Summary: My personal feeling is to train lora with a relatively small number of steps based on some face portraits, and then use lora to change the face on the basis of the picture, and the size does not need to be too large, but there is no problem in using it on portraits; you can see A lightweight face-changing plug-in
Image expansion
If you have a good picture of Wen Sheng and want to preserve the details of the original picture (or other source pictures), expand it horizontally and add background vertically, etc.,
If the prompt words produce pictures, copy them directly. The parameters are the same. For others, the prompt words will be inferred.
The original image is 1200x1200, focus on how you want to expand it, just increase the resolution
The controlnet settings are inpaint and reference respectively. Inpaint is selected at the bottom to fill in the blanks after zooming.
Effect display, only expand horizontally by 200 pixels
Photographic quality image generation
https://liblib.ai/modelinfo/4d5083a6230343869c0e940b1d5e6329
1 girl, floral dress, long skirt like smoke, petal pattern, long hair in the wind, sense of luxury, stylish still photography, looping video, captivating movement, captivating visuals, seamless blend, fascinating story, swirls, edges Clear forms, dynamic movement, bold geometry, broken shapes, sharp contrasts, modernist influences, nature, history, modern art, stunning, unreal backgrounds, soft lighting, happy moods, (near distance, face focus) (fractal art) extremely detailed, the most beautiful forms chaos, elegance, with brutalist designs, vivid colors, immersion, details in the dark
Recommended AI tools, if you are interested in playing with them
https://convert.leiapix.com/animation
Play to enlarge the video and increase the frame rate
