Yes it does work you need to chose the face ID preprocessor sdxl and the face ID controlnet model respectively if you choose the plus version preprocessor you also need the plus controlnet the only ones that don't work are the loRas for me but you are right I that I think SD1.5 has better result

いつも応援ありがとうございます！ 動画の 0:25 でインストール方法をご覧いただけます。とても簡単で、お好みのスタイルを選択するだけで、スタイルに関連するキーワードがプロンプトに追加されます。簡単ですよね、 ぜひ使ってみてください！動画へのご視聴もありがとうございました。

Thanks, they just added it in BTW the added the instant IF implementation just after I published the video I thought it will be some days but the Devs on the controlNet repo are too dast😂❤

@@impactframes Totally😁. Controlnet team is awesome & also crazy. They are updating it every single day in some way or the other.

yes you should be able on the same way you do with other models but you might be also able to lad them from HF as other CN models

Yes it is probably will do a video about it

@@impactframes Great thanks to you, I see there many people solving this problem , however no solution yet discovered. Will be grateful to you if you can find it and make a video.

Thank you 😊

You have to navigate to the custom_nodes folder and install Insightface from the requirements.txt since many nodes use Insightface it might be that you already have it and is some other node conflict try just installing the versio' from the requirements.txt or use the --upgrade flag

Did you update the controlnet extension? At the extension tab check for update and restart SD web UI? the model needs to be on your controlnet model folder. Normally A1111 will install all the necessary dependencies before start you need to restart SD web UI.

@@impactframes thanks, i fixed it!!!

Thank you for your feedback. Getting good results with the face adapter depends on various factors one I think I mention on the video is the source fotos work better if they are all from the same day people get face surgery, change hair style, get older so if the pictures have different face features is hard for the model. The type of shot is also important as you mention a face portrait is the best for this because the model loose some of the accuracy on full body shots. The base model itself. I Think you are right but not many people don't know all of the effort that has to be done to get consistency with other methods I thought this was easy and good considering the amount of effort. I really appreciate your comment for me is super hard to show my face and do the video but I am working on a new avatar that is going to hopefully feel more unique for my channel and more professional and friendly. Thank you 🙂

I wish my machine was more powerful to record stuff live but sometimes I have to show just results. I promise you all was exactly as explained I don't hide anything. It might be that you missing a dependency or have different flags or using another extension that it is conflicting there are too many variables outside the UI settings we can see that's the hard part about being early on this gen ai stuff

@@impactframes yeah I'm not blaming you haha, woulda just also liked if you showed then settings a bit slower. I think the issue is that my model for some reason isn't compatible with ipadapter. does it work with 1.5 as well as 2.0 models? or am I just completely off here?

@@spaceteapot there are two versions SD1.5 and SDXL no 2.1 yet unfortunately but tbh I only use 1.5 and XL the thing is that there are LOras too the loras don't work at all and the IP adapter needs to be pair with the right preprocessor so if you select a plus model you need a plus control model too for example for the control model named ip-adapter-faceid-plusv2_sdxl.bin you are going to need and IPadapter plus face id sdxl preprocessor if the name has no plus the you select the ip adapter preprocessor without plus in the name

@@impactframes hm yeah I made sure they're both plus. also have an sd1.5 model pretty sure. but thanks that's helpful, I'll double check again!

Once you have the preprocessor and model correctly selected, it should automatically download a dependency model when you click generate. If it doesn't it isn't gonna work. It might do you well to delete the entire installation and get something like Stablity Matrix and use that to reinstall AUTOMATIC1111. I have had better success with these things after I did that.

I used to have one that worked on colab but didn't google banned gradio and stable diffusion? I will look around to see if it still works when I get a chance thanks.

This is a free avatar from Vdroid studio, is free and you can download it. I didn't made it.

Yes we can with ipadapter plus and selecting the same models we use here .

@@impactframes thank you ❤️

Depending of the model it gets closer but the character does transfer the features instantID was just added after I publish the video so now should be even better