IP 어댑터 버전 2 설정: 설치 및 사용 가이드

IP 어댑터 버전 2를 설정하려면 comi 윈도우 포터블의 모델 설치 및 사용 가이드를 따르세요.

### ComuEye로 IP 어댑터 및 Face ID 설치 및 사용 방법
* 💻  이 가이드는 ComuEye에서 IP 어댑터 버전 2 및 관련 모델을 설치하는 단계별 지침을 제공합니다.
* 🔌  IP 어댑터 설치에 필요한 모델과 폴더를 다운로드하고 정리하는 과정을 설명합니다.
* 📷  이 가이드는 또한 프롬프트에서 사실적인 출력을 생성하기 위해 ComuEye에서 IP 어댑터 통합 로더와 Face ID를 사용하는 방법을 보여줍니다.
* 🖼️  고해상도 이미지 사용 등 결과물 최적화를 위한 팁을 제공하고, 다양한 모델 및 설정과의 프로세스 호환성을 강조합니다.
* 👩‍💻  이 가이드는 ComuEye에 대한 추가 지원과 정보를 얻기 위해 Discord 커뮤니티에 가입하라는 초대로 마무리됩니다.

are you ready for this much power? Hello, my friends. How are you doing? IP adapter version two has been out for a couple of days. It was a pain to install this, so I hope with this guide I can show you an easy way to set everything up and get it running in comu ey. Let's get started. So first things first, there's another video by the creator of these notes. His name is on YouTube, latent vision, and then also Mato. So please pummel him with his questions if your version doesn't work, and also check out his channel and this video. To get started, you need to go into comu I, and you need to go down here to manager and click on up dat. All this is going to update comu ey and all the notes that have been installed. Now, if you don't have that note at all, you need to install it fresh, of course, and that's also going to be here with that TB page. Or another way, of course, is to go again to the manager up here to install custom.

nodes, and then type the name of this extension and install it from here. That's probably the easier way to do that. The name of this, as you can see up here, is confi IP adapter plus. You want to scroll down to the area where it says installation on the GitHub page. And here, very important, it tells you what path you have to put the models into. So you need a whole bunch of models for that. You need the confu I models clip Vision. These two, now these don't come with that name. You have to rename them as it says here to this name. You can see right here for both of these files. You have to have also the comi models IP adapter model. So this goes into this IP adapter folder as named here. In case you're still confused, you're going into your comi windows portable folder. In there, into the comi folder, and in there, into the models folder. And in there, you either see these folders like clip vision and IP adapter, or you have to create them. So like I said into the IP

adapter folder download all of these models not this one because it says deprecated which means it's no longer relevant. Then you also need to have inside phas. Now this is the same ominous inside phas where there was the huge copyright issues in the past so this means the things you can create here you can't use commercially unless you have a license. I was told the license is pretty expensive you have to check that out for yourself. And then below that you also have here the IP adapter phas ID model mod. Again you need to download all of these models into the IP adapter folder not the ones where it says deprecated. And then you also need to download into the Lura folder that's also inside of your models folder these models here not the deprecated one again same thing. Now for the inside face you might have a problem running this so I want to show you how to do that. Here is another GitHub page and in there you have these different inside face Wheels as they are called. Now you need to choose the one

that has the same Python version that you are using for your comi portable. There's an easy way to find that out. You go to your comi folder to the Python embedded folder, and then you're looking for the Python x file. Double click on that. And here on the top, it's going to show you the name Python 3.11.6. In my case, might be different for you. So you want to download this file if you have Python 3.11. Now here's another issue. You might have a problem with the environment path for your portable Python install. So you want to go on your Windows search down here, type environment, and select edit environment variables for your account. When you have that, look for path, and then click on edit. It might be that here you only have one line. If you only have one line, click at the end of the line, write a semicolon, and then type anything else, and hit okay. Afterwards, go again to path edit, and you should have this list here. Now in here, I put the location of my portable Python embedded.

and just to be safe, I put the scripts folder. And also just the embedded folder. So put both of them in here. And also keep the normal python address in here. Click on Okay. Click on okay again. Now we need to start the install process. For that, you want to go into your comu ey Windows portable folder where you see the comu ey folder, the python embedded folder, the update folder in here. Right-click and then click on open in terminal. This is going to give you this CMD or command window. Now in here, you need to write this: python embedded python.exe minus M pip install then the location of the file and then onnx runtime. Hit enter and pray to the gods of AI that you don't get an error message and everything runs well. If you get an error message, best you look online in different discords for an answer because it's going to be individual for your solution. Also, in case you're wondering where do I get the path of that file, well here is the solution. After you downloaded the file.

right-click on the file and select Copy as Path. This is going to give you the path, and then you can use that in here in the command I showed you before with copy-paste or Control V. All right, so after all of that has been done, hopefully everything runs for you. And now I'm going to show you how to use this with a simple build. So we have here the classic build where we load checkpoint, we have our positive and negative prompt, we have here the K sampler, the VAE decoder, and of course our image output. Now, of course, we need the IP adapter. Now, this is different than in the past because now we need the IP adapter unified loader. This is supposed to make things simpler for you because here you have a selection of six different options. This is going to load the models for you in the background for everything it needs, so you just have to select one of these choices, experiment around with them on what they do. By the way, getting this note here is pretty easy. Double-click on the background anywhere and type unified. This is going to give

you the IP adapter unified loader and the IP adapter unified loader face ID because the face ID has a different note so as we can see the checkpoint goes into the IP adapter unified loader and then the IP adapter unified loader goes into the IP adapter here you also need to connect the IP adapter and you also need to connect the image you want to use and all of that with the model output goes into your K sampler and that is already it using it is really not hard this made using IP adapter a lot simpler and one really cool thing you can do with this is search for a clothing item online it is best if it is neutral like this with a white background and then you want to write of course your prompt woman wearing a pink puffy coat with fur Hut you want to set up here the strength of your IP adapter experiment around with that here I have a weight of 0.7 and then you get this amazing output where she's actually wearing the coat how cool is that of course you can do a

lot more with IP adapter. Also, one more thing that is important here to point out is if you switch between XL models and 1.5 models, you don't need to change anything here. You only need to change the latent size of your image, and the model for the IP adapter is automatically being changed for you. Now let's have a look on how to use face ID. It is basically just as simple. The only difference here is that you need the IP adapter Unified for face ID. You select down here one of the options you want to use. Here you can see that one of these options is only for SD 1.5. The other options work also with SDXL. Again, the model goes into this and comes out of that as well as the IP adapter here. In this case, you also have a value for the lure strength. In this case, I set it to 0.6, and the provider in this case is C CPU. Mato says to keep the CPU setting because it otherwise loaded is into the vram. It might be a problem, but you can

Certainly, also try to use the Cuda option, or any other option in here, to see how well it runs for you. Then, of course, again, this is going to go into the IP adapter again here. Play around with the weight, and we have here our amazing output for that image. Now, I found this works a lot better if you have an image that has a higher resolution. So, if the resolution is very, very small, and also the image is blurry, it's not going to work as well. A sharp image with a good high resolution, like for example 1,600 by 1,600, should give you good results. Now, here's also the very nice part about all of that. This all works with lightning models. In this case, you can see here I'm using six steps. I'm using CFG scale one sampler DPM Plus+ sde with a scheduler of normal. Again, play around with these settings to see what kind of results you get. If you want also, you can join my Discord. I've linked it below. We have a

special room just for com VII the community there is super helpful thank you for watching and see you soon bye oh you're still here so uh This is the End screen there's other stuff you can watch like this or that's really cool and yeah I hope I see you soon uh leave a like if you haven't yet and well um yeah

IPadapter Version 2 - EASY Install Guide

Olivio Sarikas, an AI Expert and passionate Artist, invites you to explore the exciting world of AI art. Join live streams to turn your creative visions into reality. Let's create together! 🎨🤖🚀

Olivio Sarikas is a professional Designer from Vienna Austria, a passionate Artist for over 26 years. He studied at the University for Fine Arts in Vienna and Munich and holds a Masters Degree in Fine Arts. 

Community Guidelines:
Inspire, Engage & Help: Use comments to share passion, inspire, and support AI art.
Respect and Kindness: Treat all with respect, regardless of opinions or backgrounds.
Constructive Criticism: Offer constructive feedback; avoid overly negative comments.
No Harassment: Report and avoid harassment or bullying.
No AI Haters: Spread love and passion; keep comments positive and inspiring.

*Note: Do not use the email below for questions.


Olivio Sarikas

IP 어댑터 버전 2를 설치하는 것은 어려울 수 있지만 다음 단계를 따르면 더 쉬워질 수 있습니다.

comuey와 설치된 노트를 업데이트하거나 TB 페이지에서 확장 프로그램을 새로 설치할 수 있습니다.

또 다른 옵션은 관리자로 이동해 확장 프로그램의 이름을 입력해 사용자 지정 노드를 설치하는 것입니다.

핵심은 confuI 모델 및 IP 어댑터 모델을 포함하여 필요한 모델이 GitHub 페이지의 지침에 따라 올바른 경로에 배치되었는지 확인하는 것입니다.

모델 및 폴더 다운로드 및 관리 지침

더 이상 사용되지 않는 모델을 제외한 모든 모델을 IP 어댑터 폴더와 Lura 폴더에 다운로드하세요.

내부 페이즈 모델은 과거 저작권 문제와 잠재적으로 비싼 라이선스가 필요할 수 있으므로 주의하세요.

인사이드 페이스 휠을 다룰 때는 코미 포터블의 파이썬 버전과 일치하는 버전을 선택해야 합니다.

또한 잠재적인 문제를 방지하기 위해 포터블 파이썬 설치의 환경 경로를 다시 확인하고 수정하세요.

Python 임베디드 설정 및 ONNX 런타임 설치하기

ONNX 런타임을 설치하려면 명령창에서 Python 임베디드 폴더로 이동한 후 "python.

exe -m pip install" 명령 뒤에 ONNX 런타임 파일의 위치를 입력합니다.

오류가 발생하면 시스템에 따라 다를 수 있으므로 온라인에서 해결 방법을 찾아보세요.

설치 후에는 IP 어댑터 통합 로더를 사용하여 사용 가능한 6가지 옵션 중에서 선택하여 더 간단한 모델 로딩 프로세스를 진행할 수 있습니다.

파일 경로에 액세스하려면 다운로드 후 파일을 마우스 오른쪽 버튼으로 클릭하고 '경로로 복사'를 선택합니다.

IP 어댑터 통합 로더 및 Face ID 사용

IP 어댑터 통합 로더는 IP 어댑터 사용을 간소화하여 이미지 및 모델 출력을 K 샘플러에 쉽게 연결할 수 있도록 합니다.

온라인에서 의류 품목을 검색하고 IP 어댑터 실험의 강도를 조정하는 데 사용할 수 있습니다.

5 모델 사이를 전환할 때는 변경할 필요가 없으며 이미지의 잠복 크기만 조정하면 됩니다.

Face ID용 IP 어댑터 통합과 함께 사용되는 Face ID는 통합 로더와 유사한 방식으로 작동하며, 다양한 SD 버전 및 미끼 강도 값에 대한 옵션이 있습니다.

잠재적인 로딩 문제를 방지하려면 공급자를 C CPU로 유지하는 것이 좋습니다.

고해상도 이미지 처리 최적화를 위한 팁

더 나은 성능을 위해 쿠다 옵션 또는 기타 사용 가능한 옵션을 사용해 보세요.

고해상도 이미지일수록 모델과 더 잘 작동하는 경향이 있으므로 좋은 결과를 얻으려면 최소 1,600×1,600 해상도를 목표로 하세요.

단계 수와 스케줄러 유형 등 다양한 설정으로 실험하여 최상의 결과를 얻을 수 있는 방법을 알아보세요.

이미지 처리에 대한 더 많은 팁과 지원을 받으려면 Discord 커뮤니티에 가입하는 것도 고려해 보세요.