Real Time Camera to AI in ComfyUI

[@enigmatic_e]

Topaz Labz BLACK FRIDAY SALE: topazlabs.com/ref/2377

[@Vsmphoto]

Oh man, I could have used this last weekend. I did the same thing with Touch Designer and Compute Render, but this looks way faster and has custom models, which is crucial. You are great, and I love watching your videos.

[@enigmatic_e]

🙏🏽🙏🏽

[@digitalfun2072]

Wow, I am really wondering if TouchDesigner could send and receive the real-time image to ComfyUI. Sounds like a AI camera interactive installation project

[@dbusmire]

Lmao same!

[@Vsmphoto]

@@digitalfun2072 I was just looking into this. Were you able to make any progress getting it to output to touch?

[@louward7823]

@@digitalfun2072 DId you ever try this out? or get a better frame rate than the above example?

[@gabrielmoro3d]

It’s absolutely insane how fast this is 😮

[@enigmatic_e]

I know right! It’s going to be crazy when this could run without any lag or frame skipping.

[@rodneyjweltham150]

That is fascinating. I noticed that the processing is really slow. I would imagine you need a quantum computer to make that thing run smoothly, nevertheless it's interesting to see what can be done

[@ParvathyKapoor]

Wonderful! Thanks for the tut

[@enigmatic_e]

Thank you!

[@AscendantStoic]

We are very close to having older games being remastered through rendering in real time using AI, it will be glorious 😊

[@enigmatic_e]

That’s going to be crazy!!

[@CouchstainCollective]

What would be cool would be to turn this exact concept into a VR application

[@mohammedrashid1910]

damn this so cool. ai is now getting close to allowing content creators to just turn a camera on and make a cartoon in realtime. cant wait

[@JingIeFett]

I'm having issues opening CaptureCam. I did the git install, clicked setup bat and then after when I click run bat, a command window briefly opens and closes and then nothing happens. Any ideas?

[@BrunoPiresBPS]

same here

[@lunarlouie]

@@BrunoPiresBPS do you find a solut8on to this?

[@lunarlouie]

did you find a solution to this

Nice :) Though this still needs to become much faster so we can say it's real-time. Maybe this would be a great candidate for the TensorRT compatible SDXL to show it's muscles. Nevertheless, it's awesome we have so much toys to play with.

[@Duckers_McQuack]

Though, one node we do need, is one that can create a virtual webcam being those generated frames :D

[@normalplayer163]

Great Video! Easy to understand! but I got stuck on the step where I have to run bat. I tried several time and that webcam app didnt show up. HELP PLEASE!!!

[@KolTregaskes]

Very cool. Have you also seen Krea? Someone has hooked that up to a webcam too and is basically the same thing, without ComfyUI.

[@enigmatic_e]

No I haven’t. I want to check that out.

[@jaharoni]

Quick with it!! Thank you for your dedication to these tutorials

[@marcjoymedia]

Thank you so much for sharing. So many new possibilities!

[@PedroKozovitsBarbosa]

i DON'T HAVE THE "LOADwEBcAM" NODE

[@MahmoudRadwan-j4n]

you can batch those generated single images photos into animatediff node or animate diff combine and create a gif or a video. Nice tutorial thank you :)

[@TR-707]

thanks for your video! Was looking for a solution for this. My plan is to just pipe a virtual cam/ desktop capture into comfy and this is a simple solution

[@enigmatic_e]

Awesome!

[@JustMaier]

“This is if you wanna catfish somebody at some point” 😂

[@enigmatic_e]

😂 you know its gonna happen

[@TheAgeofAI_film]

Thanks for sharing, I have subscribed, this is very interesting, I love to try this in our short films...

[@Vsmphoto]

Is there a way to output your preview to another screen or to syphon?

Great video, thanks !!

[@HiddenInformation10]

It works, but I have a question, could those photos somehow be made as a camera, for example for obs?

[@pavpav8627]

Was just thinking about how to set something like this up and boom - here it is !

[@enigmatic_e]

It was bound to happen

[@JeremyBotz]

I can't get past the computer vision init. It keeps saying cv2 module not found, i've confirmed openCV was installed. Need help

[@CineFlavor]

My LoadWebcamNode doesnt give me an option to load the image path. It just is blank and red. When I first drag the image of the node layout on for the preset it tells me there is no webcam. So all I have is a blank node with a title. What is this type of node called? Can I add it manually?

[@enigmatic_e]

Have you updated everything already?

[@enigmatic_e]

If you are on webcam 0 try webcam 1 or webcam 2.

[@CineFlavor]

There is no option to choose webcam 0 or 1. Its just a blank node. Everything should be updated since i just installed it last week. My webcam is plugged in via USB so i'm not sure if that changes anything. it's the Elgato Facecam@@enigmatic_e

[@CineFlavor]

The issue turned out to be that I needed to install the missing nodes using ComfyUI Manager@@enigmatic_e

[@Dante02d12]

The little app is confusing. It seemed to work at first, then stopped working. It created a new render image by itself first ,then I reinstalled it and it doesn't seem to do it anymore. I just don't understand how to use that thing. Any advice? It's super interesting for so many things!

[@juschilltech]

Yall are KILLING your ssd

[@wuetsby5448]

this is cool

[@aimademerich]

🔥 🔥 🔥

[@enigmatic_e]

🙏🏽🙏🏽🙏🏽

[@Iamdevvvv]

We are now more closer to see whole world through ai😅

[@enigmatic_e]

I wonder what AI will look like a year from now.

[@koosiuhang5257]

where can i get the ModelSamplingDiscrete node , i tried the install the missing node , but not working , 🙏

[@enigmatic_e]

that should be a native comfy node. Have you tried updating to the latest version of comfy?

[@koosiuhang5257]

ya , i did manuel update though download the comfy ui zip file and copy and paste the extra node folder and work ! @@enigmatic_e

[@nickdivito4662]

Can you combine something like these new LCM models into Vid2Vid work flows to get faster render times?

[@enigmatic_e]

Yes you could!

[@louward7823]

Did you try this out? or had any luck with better frame rates?

[@InkaHacker]

Do you recommend using this on Linux or Windows? Would it be too much resource intensive for a RTX3060 while streaming?

[@enigmatic_e]

I don’t think this is ideal for streaming, at least not now.

[@USBEN.]

Damn bruh real time deep fakes, oooo baby!!

[@enigmatic_e]

🤣

[@genAIration]

Sir how do you show up your embeddings? I mean like a pop-up one. I don't know this node

[@enigmatic_e]

When I start typing embedding in the negative prompt, it automatically shows me embeddings I have available and I just select what I want.

[@genAIration]

@@enigmatic_e oo I see. I use a lot comfy but always use sdxl models but i don't have any sdxl embbebings then thats cz its don't show up to me . Thx so much sir 🙏🏿

[@shopbc5553]

I would want to output HD 30-60 p or even 4K 30; is it possible, and how big of a computer would I need to make?

[@enigmatic_e]

I have a 4090 and I dont think I could do that with this gpu so not really sure.

[@jbiziou]

very cool,:) I got stuck at ,, ( Load WebCam Image Node was RED warning) I then ran the run.bat in the CaptureCam folder, and restarted but nothing worked. Launched OBS and started recording, and restarted Comfi, but still not working. Any Idea please ? thanks :)

[@enigmatic_e]

Not sure what it could be. I would go on X and message the developer, he said to reach out to him if you run into problems. I put his link in the description

[@jbiziou]

@@enigmatic_e Thanks:) will give that a try.

[@jbiziou]

Ahhh think I got it. In the manager I loaded custom node for Webcam, and boom, the node is no longer red :) got the preview image in the webcam,, but its squashed, haha,, Progress though:) 1 step at a time

[@jbiziou]

Last thing was ,, Had to rename the LCM lora to ( pytorch ), from LCM.

[@TheHologr4m]

so my little i7 laptop with GTX1060 3gb probably stands no chance at behaving anywhere close to your monstor of a machine.

[@frankyaKaBIGMAC2]

u sir r a fcking gangster og

[@enigmatic_e]

🤣 thanks!!

[@koningsbruggen]

Can I use a screencast as the output?

[@enigmatic_e]

Haven’t tried it, might be possible.

[@hiddenintheshadows530]

screen cast? It has Region Capture, you can capture your screen and feed it into ComfyUI. Once again @enigmatic doing only part of the full tutorial

[@enigmatic_e]

@@hiddenintheshadows530 my bad I’ll try to do better. I’m still in the learning process. 🙏🏽

[@liquidmind]

webpage is gone?

[@liquidmind]

This site can’t be reached

[@enigmatic_e]

@@liquidmind not sure, things in the AI world changes so much it’s hard to keep track of it all unfortunately.

[@Parzival047]

.....so many things missing :(

[@enigmatic_e]

What is missing? I can try to include it in the description.

[@Parzival047]

@@enigmatic_e bro I went on to install comfy UI, then I downloaded that aniverse, but how to add this aniverse into comfy UI? I try loading it but it just doesn't loads. At this point, I am just installing every model which is there 😭

[@Parzival047]

If I follow your path to install this... Then. 1. Install comfy UI 2. ? 3. ? Then what should I do next such that it WORKS 😢

[@enigmatic_e]

You save that model in your checkpoint folder. Are you new to comfyui? I have a tutorial on how to get started kzbin.info/www/bejne/jXnbeqWPZ9x_h6c@@Parzival047

[@Parzival047]

@@enigmatic_e yes bro new to comfy UI ❤️

[@jasonkaehler4582]

run.bat just chokes, says old python, but have newest and hand-installed everything the toyxyz 'requirements' doc. any suggestions?

[@kazrik257]

I installed python 3.10 and 3.11 through the Windows store and it worked

[@mick7727]

Lcm always looks trash for me. Must be the models.

[@enigmatic_e]

Yea might need some tweaking of the workflow too

[@kachuncheng-s1v]

thanks~! Is it possible to use the capture cam if I run my ComfyUI on Runpod?

[@enigmatic_e]

Sorry not sure. I never ran it with Runpod.

[@KennethEstanislao]

What’s your gpu?

[@enigmatic_e]

It’s in description. It’s a 4090

[@KennethEstanislao]

@@enigmatic_e nice! sorry I didn't saw it as some youtubers usually didn't put it anywhere! thanks for the reply!

[@enigmatic_e]

@@KennethEstanislao all good. I recently just started doing it because many ask about. 👍🏽