HIghly experimental concept for girl idols concert/music show.Extra inpaint works will be needed to assign better and different faces for each idol. I'm still experimenting finding best prompts and settings. Please try it yourself and give me feedback to improve for next version.Trigger word: idol stage Possible words: multiple girls, 3girls, 4girls, 5girls, 6+girls, dancing, singing, microphone, crowd Sometimes exactly right number, sometimes just ignore the numbers. Restore faces will need to be ticked on unless you want to do faces inpaint later then its not necessary Update 1: I noticed that setting bigger picture will give better details on limbs and face. In the 1st picture I used 768x512 with hires fix R-ESRGANv3 at 2.5. I can't go any bigger than that with my puny 10gb GPU. Need help to test how it will look like in a full 4K resolution image with 24gb GPU. If you guys do try, please post it under this model or inform me somehow. Thank you.Update 2: I found this tool that basically does automatic inpainting with automatic face detection. <a target="_blank" rel="ugc" href="https://github.com/Bing-su/adetailer">Bing-su/adetailer: Auto detecting, masking and inpainting with detection model. (</a><a target="_blank" rel="ugc" href="http://github.com">github.com</a><a target="_blank" rel="ugc" href="https://github.com/Bing-su/adetailer">)</a>. Unfortunately adetailer doesnt support scripting to feed different models for each found faces. This can be done by running the generation n times for the n girls in the image using different models each time, using same seed in the main generation and everything. Then use photoshop to crop each faces to merge into 1.

Files with SHA256: 735d0287cf49ed722d1874f22dc2c81a674d3bc061c3f96d465095bb8adc522b