AiPlaygroundWM

 Detail deamon extension test Check Point : MagicaPony realism darkMagica xyz plot Amount and mode test 1st 1. 2. fail detail deamon off so doesn't apply 3. Amount over 0.4 starting looks crazy. remove over0.4 4. Hm... amount 0.3 looks weird. 5. Hm....LoRA may be causing some ambiguity in the test. remove lora. add amount 0, -0.1 6. 7. In amount 0 each mode has same result.(same as off extension) Amount over 0.2 more saturated and weird.(not more realistic) Minus amount and 0 amount remove in this test. Conclusion1 : Mode cond and both looks fine. (It is unclear which one is superior) Amount over 0.3 start looks weird and too saturated Mode : both Amount 0.1, 0.2, 0.3 and  end 0.6, 0.7, 0.8(default), 0.9, 1 test 2nd 1. Remove amount 0.3 2. Amount 0.1 start 0, 0.1, 0.2(default), 0.3, 0.4 and end 0.6, 0.7, 0.8, 0.9, 1  test 3rd 1 2 The test design seems to be faulty. After testing the detail daemon, I'm unable to establish a clear direction. Its use will be temporarily...

In the middle of testing to find out settings values. I have an issue on SDXL and pony . If you look at it enlarged on PC, you can see hexagon texture skin, surface and background. Every single image have it. Changing quality prompt, image resolution, VAE , CFG, steps , not works at all. " AH!!!!!!!!!!! " But today, I found out how can fix it!! The answer is Tiled VAE turn off while using SDXL and PDXL  ! Sometimes my laptop cant't handle SDXL and pony.  Using Tiled VAE could help. So I always Tiled VAE turn on. " Satisfying "

About 1 year ago I've tested denoising, and today test again.   New ram has arrived~ 8GB *2 ==> 32GB *2 upgrade! 15steps 512*640 (4:5 ratio) hires 2.5x Check Point model is :  AstrAnime 5b 0.4 0.5 0.55 0.6 Tiled diffusion extension off by mistake, but runs well!? Keep off and tested. [ The face all good because using AD(after detailer. faceand hands. denoising 0.55) ] 0.55 and 0.6 slightly different. and better than 0.4 certainly. Similar results. Change Check Point model : meichidark V3.8 ????? Same as before How about 0.7? ????Ok. Pretty mess up Remove 0.4 0.5 0.55 0.6 Denoising over 0.55 is better than I think. How about at steps 20 0.5 0.55  0.55 result's heels started wierd. How about with lora? 0.6 Ok definitely 0.6 is out. Conclusion : Denoising is under 0.55 Every image has exif info

Using : Negative Guidance minimum sigma (skip negative prompt for some steps when the image is almost ready; 0=disable, higher=faster)  Settings DPM++ 2M Karras 512*640 resolution hires 1.5  batch count : 4 Adetailer on (face, hand) Sampling steps: 15  Lora included Check Point model :  majicmixRealistic_v6 After every test group, reboot laptop Negative Guidance minimum sigma : 0  (default) test result (time)  1. 3min 1.6sec 2. 3min 3.3sec 3. 3min 3.2sec 4. 3min 4.3sec 5. 2min 58.4sec 6. 2min 56sec 7. 2min 51.5sec 8. 3min 11.6sec 9. 3min 0.8sec 10. 2min 53.4sec 11. 3min 3.4sec 12. 2min 46sec 13. 3min 14. 3min 5.5sec 15. 3min 3.8sec 16. 2min 59.3sec 17. 3min 0.4sec 18. 3min 10.7sec 19. 2min 46.4sec 20. 3min 1.3sec Avg(Average): (except longest and shortest time, simple average): 3min 0.2sec (Round to the nearest second decimal place) Negative Guidance minimum sigma : 0.1 1. 3min 1.1sec 2. 2min 52.2sec 3. 2min 52.3sec 4. 2min 52.4sec 5. 2min 57.8sec 6. 2min ...