Amazing updates for Hires fix and SDXL faster refiner pipeline in A1111 v 1.6 for stable diffusion

[@davidbene9962]

The best part in your videos is when you say "And have a good day!" Not because its ending, i love your videos. But the way you say it so suddenly, always brings a smile on my face :,D

[@AI-HowTo]

Haha, sounds robotic doesnt it :), and thanks for the cool comment.

[@ParvathyKapoor]

Thanks for the update

[@AI-HowTo]

You are most welcome.

[@sb6934]

Thanks!

[@AI-HowTo]

You are welcome.

[@coloryvr]

...a great update! But Deforum is gone now! Will it come back?...I hope there will be an update for Deforum soon??? Happy colored Greetinx!

[@AI-HowTo]

not sure if its gone or not, since the extension is still there, even in A1111 v 1.6 ... if dev stopped then probably for a good reason, such as presence of practical alternatives

[@AI-HowTo]

Just double checked, it is not working indeed in V1.6 and several other extensions ...current solution i guess is to downgrad, you can watch this video if you dont know how already kzbin.info/www/bejne/hpXOc6B6pLp8r5o

[@coloryvr]

Super! Thank you so very much! Happy colored Greetinx!

[@coloryvr]

❤ @@AI-HowTo

[@AI-HowTo]

You are most welcome

[@razagr]

Thanks ,a tutorial about how to have consistent character with same clothes and background , but just changing the pose?

[@AI-HowTo]

this is achieved by Controlnet, I did a tutorial before about Controlnet which could be used for this purpose kzbin.info/www/bejne/Z2TJmHV4fpKNjrc ... for example, you can use After detailer to have same face all the time, Controlnet to control the pose, same seed to have same clothes ... may later do a short video on this subject regarding characters only as well.

[@AI-HowTo]

that tutorial was generally about Controlnet, not about Consistent characters though, and thanks for the suggestion.

[@razagr]

@@AI-HowTo thank you I tried that, even trained a lora for my character but it was not consistent face either clothes, I am sure I am missing something,

[@AI-HowTo]

achieving 100% consistency is not possible in SD since it has random seed/elements, so no matter what we do, something will remain missing ... but with the previous, it becomes possible to achieve high level of consistency, will check if i can create a video about this soon.

[@razagr]

@@AI-HowTo Thank you, that would be of a great help.

[@thedevilgames8217]

sdxl still slow for me 30 min to wait hires.. not sure if its my graphic card or its a bug, my graphics card is 1080p 32gb ram

[@AI-HowTo]

this should not happen in general, possibly a graphics card issue, I find SDXL 2-3 times slower than SD 1.5, so I still use SD 1.5, either ways if SD 1.5 works fine for you, stick with it, SDXL is not worth it unless you have a powerful Graphics card.... or seek alternatives such as running on the cloud/colab/kaggle...etc.

[@maxrinehart4177]

​@@AI-HowTo is 3070 and 4070 good for SDXL?

[@AI-HowTo]

mine is 3070, i can generate images in SDXL with refiner application too, but it takes more than 20 seconds and more than 40 with refiner too, not the best option, but acceptable enough, currently if one can affort 3090 then it is best for the long run, especially if you plan to do some training, after a while 3070 may no longer be enough @@maxrinehart4177

[@lioncrud9096]

is refiner needed if you're using a different SDXL model than the 1.0 base?

[@AI-HowTo]

no, refiner is not needed even with SDXL base model, it just improves results slightly ... it can also be used as standalone model like the base model as explained in my video kzbin.info/www/bejne/iofInKJ5p9hge9E ... and it can be used to refine real world images or images generated by other check points to improve its details and possibly alter it depending on number of steps used.

[@gleentea]

love your work. How do we get in touch with you

[@AI-HowTo]

I am sorry, trying to be out of touch at the time being, not accepting private or business inquiries yet.

[@gomezabel8880]

The Tutorial show the file name extension "bat" but my stable diffusion fold shows the WebUi-User but the "bat" is drop (missing) I'm a newb and just finally installed Auto1111 last week, would like for someone to shed light on my issue. Thank you

[@AI-HowTo]

this is because your browser settings do not show the file extension ... if you are on windows just edit the file with icon that looks like two gears ... or when you right click edit, the content of the file will start with @echo off so that you dont confuse it with the linux version file because there are two files in the same folder... then proceed as mentioned in the video.

[@Jaradis]

In Windows, in File Explorer, go to View, Show, File Name Extensions. Make sure they are turned on. Might as well turn on Hidden items while you are in there.

[@gomezabel8880]

@@Jaradis Thank you very much for that advise :)

[@asmrannoyyou1644]

Hello Bro why didn't I install stable diffusion in my laptop ? please give me some solution

[@AI-HowTo]

github.com/AUTOMATIC1111/stable-diffusion-webui following installation instructions should work... if you have low VRAM, you can use --lowvram or --medvram option in the webui-user.bat args ... if it still not working because you have lower VRAM than 4GB for instance you can try ComfyUI github.com/comfyanonymous/ComfyUI which can run even on CPU ... while Google Colab could be a better option if you GPU is not good enough... keep trying, it will work out eventually.