2D Character Animation In ComfyUI
HTML-код
- Опубликовано: 6 июн 2024
- The first 500 people to use my link will get a 1 month free trial of Skillshare skl.sh/sebastiantorres01241
The Method I use to get consistent animated characters with ComfyUI and Animatediff. BYO video and it's good to go!
Want to advance your ai Animation skills? Checkout my Patreon:
/ sebastiantorresvfx
For the companion PDF with all the links and comfyUI workflow.
www.sebastiantorresvfx.com/dow...
You're awesome! Thanks for hanging out with me!
looks amazing. thank you for shearing your ideas and workflows
I only watch the intro and Is just incredible. You are the most underrated youtube channel for this topic. Keep it up and you'll be one of the top 10 soon!!!
Well that comments getting printed and framed! Haha thank you Javier you’re awesome! If you guys keep watching I’ll just keep making them.
This looks like exactly what I want to learn!
Some good insights about character positioning!
Great work mate, always a good day when you've worked out something new
haha I wish it was just a day. The trial and error phase takes me way too long...worth it yes, but I'm hoping to speed that up.
I've been messing around with your workflow from last video and trying to see how much I can squeeze out of it as well. It's trial and error I feel it's so easy to break things for the worse
Oh yeah believe me we break these workflows every two minutes 😂
The faster your iterations the better. So on the image loader most times I’ll set the top number to say 5 so it only does the first five frames. Just makes the breaking and fixing process move along faster.
Love you ,you are simply great.❤❤❤
Noob player, you’re awesome! 😀
best tutorials on 2D ai animation
Thank you 😁
What's your VRAM? 12 or 16 gb? Or what would be the system reqs for this?
Impressive, thanks. Will this work also for lips sync to a speech ?
We’re still working on lip syncing. That’s a little more complicated because it’ll depend heavily on how well trained a Model or Lora model is. It’s definitely on the cards for a future video though.
Would looooove to see this! Lip sync is through Canny would be the last step, otherwise, it's just fun visuals. @@sebastiantorresvfx
I’d like to have both 😆
Impressive, Thank you! but i have a question! is there a way to add an image as a Reference for the Character?
Ipadapter is the way to go. You can use as many references as you like to drive the image
ControlNetLoaderAdvanced, LoadImagesFromDirectory can't install those 2 nodes
Very interesting, thanks.
Is there a way in comfyUI to make a 3d object integration in a video matching both the video lighting and the video style?
No one seems to be capable of until now.
Not yet but I’ve seen some 3D integrations coming soon so hopefully it won’t be too long away.
great!!
@@sebastiantorresvfx
When you cranked the lighting up, I just had flashbacks to BadTech - the best online comic ever created... fight me ;) haha
It’s the eyes isn’t it? 😂
When I try Installing the missing custom nodes, It says (IMPORT FAILED) when I install them, I don't know why. do i need to install all 538 custom nodes?!
You may need to install those ones manually but first I would uninstall them from the manager then go into the folder and make sure there’s nothing of that specific custom node left. And try the manager again. This happens to me every now and then and yeah it’s annoying to say the least
I would love to discuss this with you, do you have any one on one time available on patreon?
Hi sorry for the late reply. What did you need to discuss? Reach out to me on Patreon if you like.
Something is wrong with the newsletter, the button is not submitting a form. Please check your website to fix it
Check it out again should be working now
Which human generator are you using here?
That’s the name. Human Generator. Its in the blender market website. It’s a paid addon but You can also use DAZ studio which has great models and is free, as if Unreal Engine.
you don't have a Sign Up To The Newsletter button on your site
It’s up and running again 👍
Really don't understand why more 3d artist aren't doing this more with AI. AI then just becomes another render engine or filter of sorts that when applied correctly and can correct a lot of mistakes or shortcomings of typical 3d imaging .They compliment each other nicely by offsetting the shortcomings of the other, i.e., bad hands, character inconsistency in AI can largely be corrected or mitigated with 3d input images and bad, hard to get photorealism of 3d can be greatly enhanced with AI. Text to image/video can only go so far at least for now.
Mostly because you don't have enough control over it.
This video for example, use a different Thumbnail to clickbait me into playing it, while the result is in a different style.
Ego... Many artists hated it because now others can become artists without going through the hardships they experienced. It's similar to how canvas painters felt when Photoshop was first introduced.
@@nufh I think it's not ego. 3D artists I know usually don't care about it as much as 2D artist (well, doing 3D is already considered "cheating" by some).
Whether it's true or not, the problems still stand that the result is not as controllable yet.
Mainly because the end result is not as predictable as when you do it by hand or modeling, and outsourcing the fine expressiveness of art style to an external neural network disconnected from the artist's life is quite limiting.
But! An A.I. that renders 3D is quite nice, we just need fine control over it, and hopefully in a way that doesn't feel like drawing with a microwave. :S
As many have written before. Lack of control and fine tuning. If you are a skilled 3D artist it’s actually slower and also require new learning and setup.
I spoke to some 2D 3D Fx artists and there stance is like this:
The AI tools is evolving rapidity. I rather work and focus on my workflow now. In 1-3 years when the tools are more mature I jump in. A lot of the effort learned in specializing tools using now by many AI artists will be wasted. Tools will be easier more mature with more features and most of all more predictable. I’ll jump in then an. Wipe the floor with any AI artist that lack deeper lnowledge of other tools and most of all often lack experience and training in core art / anatomy / lighting
I fully understand and agree with their reasoning above
If you use unreal engine, your rendering time will be close to 0. If you need help, let me know.
Actually have been thinking about learning Unreal, for the renders yes but also the mocap and face capture implementation which I’m currently in the R&D phase to implement to this current workflow.
I’d like help.
was the thumbnail AI?
Sure is, in comfyUI using Ipadapter.
wat prompts did you use @@sebastiantorresvfx
love it
Thank you 🙂
Why would you use depth and openpose preprocessors on the images when you have a 3D model in blender 😂
Because showing the other method would have been a much longer process for anyone just starting out. Also I did use blender for both initially. Results for a cartoon character weren’t great. But feel free to play around with it.