AI Animation That Changes The Game
Вставка
- Опубліковано 16 бер 2023
- Take full control of your AI actors. This is the workflow that I put together to take advantage of the Brand new First iteration of the script extension from my friend Sixhaunt:
xanthius.itch.io/multi-frame-...
Support:
🖱️Join Our Newsletter to win $25 Dollars of Runpod Credits !:
promptmuse.com/news-letter/
🖱️Instagram: / promptmuse
Twitter: / promptmuse
🖱️Kofi (Buy me a coffee): ko-fi.com/promptmuse
🖱️Discord: / discord
---------------
Timestamps:
00:00 - Coming :)
---------------
Helpful Links:
🖱️3D Character Creator Programs:
Character Creator & Iclone : www.reallusion.com/store/prod...
Free: www.blender.org/
Free: www.mixamo.com/
🖱️Mocap Data
ActorCore: actorcore.reallusion.com/
Free: app.plask.ai/
🖱️ Automatic 1111 Script
Sixhaunts Multi Frame script for automatic111:
xanthius.itch.io/multi-frame-...
🖱️Runpod Link:
tinyurl.com/4ubfeyf2
🖱️Civit AI Models & Lora:
civitai.com
civitai.com/models/4201/reali...
civitai.com/models/7468/scarl...
🖱️Civit AI:
civitai.com
I'm going to add the write-up here with all links...coming shortly!
P.S. You can run your frames through Ebsynth to make them smoother....Love ya
PPS For Arcane style characters Use AnythingV3.ckpt (Don't forget the VAE file- add it in the same folder as your model).
There are some nice Arcane Lora's on CIVITAI if you do a search....
#aianimation #aiactor - Навчання та стиль
Any common FAQ I’ll put here to help others 🎉
Thank you SO much for this!
just cuz you're cute doesn't mean I'll subscribe,, maybe.
Ahoy. How safe is it to use civit AI? I ve read, that some models could contain malicious code. Do you have any experience on that? thx in advance
@@michaspringphul //This is wrong. It's bullshit!
PIL.UnidentifiedImageError: cannot identify image file 'C:\\Users\\yorki\\AppData\\Local\\Temp\\ThirdPersonMap9n2p4740.0024.png'
i got this err...can you please help me?
This is nuts but the crazy thing, is that any day now, there could be a video that basically gets rid of all of these steps. That's how fast AI feels like it's moving.
Totally! This will be one click and done job soon 😅🤣 Someone in my comment said about controlling animation with neutral networks in the future. It’s gonna get WILD! *click* ✨
@@promptmuse I feel like one day, we could feed an AI a movie and replace every actor with whoever we want (or whatever we want, like animals.) It's just going to keep getting better and better. Also more terrifying lol.
This is such a true statement. In the last four months, the last two years I spent learning video game development has been rendered irrelevant. The future is now.
fully agree…upload a photo of the face/body of your choice and get back a fully operational replika AI ready to chat and become your companion
@@Williamb612 They can train chatbots on TV characters so if you want to hang out with Joey from friends, you'll be able to.
Using 3d char as a reference is the best option.
One of the things that break cartoonish effect and give rotoscopic look instead is micromovements.
3D usually don't have that because mocap kits are not advanced enough to record those movements without recording tons of movement noise.
I'm not to this point in my learning curve yet but I'm always excited by your tutorials. They're fantastic.
I’m glad you like them ! ✨
same , But I'm always learning stuff from watching the workflow! Ty for not skipping any steps prompt muse as a lot of guides gloss over the smaller details. I assure you some newbie like me is like how tf did they get from step a to step g?
Just enjoy other things than learning that stuff . It will be holdodeck at this pace in a few months. You say what you want and it presents it to you.
Love the hair and clothing change at 18:42 😊
Wow Amazing. You are so good at making the complex understandable. Please don't stop. You make the world a better place!
Be mindful that currently works created with Ai can not be copyright protected, various studios I've worked with have already been asking me to NOT submit any items outside of previsualisation stages for any projects. There also is already president established in court where an AI graphic novel was denied copyright protection. A ton of work for nothing essentially.
Thank you so much for your calm and meticulous explanation! And an even greater thank you for the overview of the whole process at the start - this creates a general map and then populates it with details so that people with different prior experience will be able to follow either on their own or with your kind guidance
You are always at the forefront!! Very creative and intuitive!! We congratulate you and thank you for your teaching.
Your process summaries are awesome. Thank you for all your efforts!
Hi Promptmuse, trust all is well. Just wanted to say you’re the definition of dopeness! Your videos are most instructive and insightful. They are very easy to understand and you top it all up with a pinch of humor, classic formula for success! I think your channel is going to be a space to watch. Keep bringing us all that nice content and we’ll stay glued to the channel!
Splendid video and fantastic workflow. Thank you.
Amazing work, such a clean animation, and without needing a film studio
Outstanding. Hope to see the next iteration soon.
Tip:
In after effects - the motion blur tag only works for objects that are physically moving in position/ scale, or in 3d space with camera movement. There isnt any motion blur on a static image like the one you had.
Also, Pixel motion blur must be the top effect or in a pre-composed layer, before any other effects othewise it bypasses those effects - your camera shake deblur is bypassed and isnt working.
Faceware require at least 198 blendshape and it is different skills we work with faceware, so the pipeline is different however adding the hair cinematics style is interesting over a model from 4 character, it is a great software
I am so new to this that I feel so green. Your tutorials are great. I am really happy to learn how to do this. Your presentation & demo is top grade. TQVM!! ❤❤
I am so glad I found your channel. AI is something I really want to learn how to use more & your tutorials are extremely helpful!
It’s so much fun ! Equally frustrating when things do not work. But when it works it’s very very cool ✨
That was so fun... and what a lovely model :-) So much detail!
Fantastic work I have a project I want to do over the Summer to enable my students to use Stable Diffusion in their own pipeline next year and your tutorials are invaluable in my understanding the process so I just want to say a huge big thank you for what you do it is hugely appreciated .
replacing to learn to draw or beside?
I think this year we will see lots of amazing AI tools that will make our lives easier. Those who can use these AI tools smartly will be the gainers. Recently I created amazing images with Blue Willow
Very well said. Artist will be artists whatever medium they select.
I've debated getting character creator 4. This definitely makes it more worth it. Thanks for all the hard work on these fantastic videos! 😃
This is so awesome, your such a great creator!
Thank you for your response. As for the script "BETA multirendering", it shows that the process is going on, but not always. One question: the script draws in the beginning small (2-3 pictures), and then the format that I set? If you have the same, then what is it for? But the fact that this script is the best I've worked with it is just great! Perfectly draws (animates) hands, fingers and even teeth! You just need to set the keyframe correctly when the person is smiling, and then everything will work. And most importantly, very easy to install.
This is the best workflow I have seen for producing coherent character videos.
Thank you
Wow you're blowing up hey! Quality content will do that! Must be very exciting to have your hard work pay off! Congratulations and well done. Keep up the great work. I hope to join you on my own UA-cam journey soon!
The video tutorial explains how to create coherent AI animations with control over every aspect of the characters using a workflow that involves purchasing or creating mocap data, creating an actor to guide the AI, and using automatic 111 web UI with six hunts python script and control net enabled. The speaker demonstrates how to create a base character and add features like hair and clothing with character creator. They then show how to use the mocap data to make the character talk, before rendering and stabilizing the animation using stable diffusion. The video also mentions the use of a Scarlett Johansson trained Laura fowl and diffusion checkpoint file, and the possibility of training your own to create your own styles.
amazing and spot on! in love with this tutorial
this is more like an AI filter over a 2D sequence that the AI doing the animating of the character
I think adding timestamps to a video of this length and complexity would be helpful. Thank you for the video though, I appreciate you and your work.
Great Workflow! Thanks for sharing!
Thank so much !! ✌️
Really great tutorial. Thank you so much
Really interesting method! Can’t wait to try it out.
love your tutoriais!
Hey, thanks so much 🫶
bravo! simply impressive. thank you for explaining such with comprehensive steps. joining your discord community now :)
Firstly, your experiments are very much appreciated and the footage looks amazing and I love this channel to bits. Thank you.
I'm wondering, what happens if you run the exported footage, from after effects, again through stable diffusion but lower the denoising strength substantially. Have you tried running the same footage again with "milder" settings, if you will? Wondering if stable diffusion itself can be used to remove the flickering or make it even more consistent.
Also, what if one were to nudge the frames towards the Johanson... not in one conversion... but 10 very mild conversions.
Anywho, I don't have in-depth knowledge of stable diffusion.. still learning and wondering.
This is actually genius. I never thought of using mocap data, and i didn't even know about character creator. Thanks so much =)
I know it sounds like i’m sponsored by them… I am not at all haha Just love there product. I’ve used Daz but personally doesn’t touch Character creator in my experience.
Just love that Character Creator flows into Iclone, Actor core then can be exported to Blender / Unreal / poison of choice 😅
Thank you for these tutorials!
GREAT TUTORIALS . YOU ARE A TRUE GEM IN UA-cam WELL DONE
I actually like the flickering, it's really dreamy 🐙💕 The deflicker is OVERPRICED 😠
Also, you can run videos through VLC and achieve a less-flickering video that way. I often edit videos on the fly in VLC and the simply record my desktop. 👍 Saves on rendering times and hassling with codecs. ✌️
Love your vids 👍 Thanks for sharing!
To be fair, it's not the plugin that's overpriced, you just need to buy the pro version of Resolve to get it.
How do yu deflicker with VLC, what is the option called?
i love how you explaining things !
🙌🏿🙌🏿thank you for an awesome tutorial
Your tutorials are amazing. Thankyou!
Thank you so much!
Great tutorial.I tend to use DEforum with controlnet to try making animation, but there to much variation when i keep the framerate of the video. Is using laura model help to get less variations? I think there is another tutorial i saw for image 2 image that was using batch files.
Holy hell. I didn't realize it would look this good 😮
Best work flow so far. Love it!
Appreciate it!
This is amazing. Thank you!!
Gosh, top TIER content! Thank you for sharing another fantastic workflow video!
Glad you enjoyed it!✨
@@promptmuse I'm applying these techniques as we speak on my anime model. Thank you!
@@activemotionpictures Let me know how you go!!
@@promptmuse I'm tagging you for sure.
Best most helpful, informative, creative, educational and inspiring UA-camr I ever watched. I mean it from the bottom of my heat.
If you have a Patreon or something like that, please share the link I NEED to support your work.
Any idea as of yet on how to fix the hair , however? It really gives it all away.
Great video. Thanks for sharing your knowledge.
Thank you for your comment 👌💕
great work.. I'm not ready for this but I really enjoyed the video plus the tips I got from you for my Leonardo images, Thanks!!!!!!!!!!
Great to hear! Thank you 🫶
Love your videos! would love more tutorials on character creator if you had time x
Yes got one in the pipeline for sure !!
Yooo! Somebody send this to Corridor Digital and The Russian Badger! This will flip their heads.
😂😅🤣
Thanks! Amazing tutorial.
WOw! Thanks so much for the support!!!
Really well done videos!
Love this, Thanks for sharing .. 😊✌🏻
As a gamedev I'm eager to see neural networks to simulate 3d movements.
Wow, this would be mind blowing 😮
@@promptmuse you are amazing. Thank you for teaching me so much.
🙋♀️ what cloud system is best. Can it all be on run diffusion in the cloud or should I learn the Pod thing. My computer capacity is so small it won’t load a browser page 📄 😱
Also Can I borrow the wigs 😂. Just kidding
Cascadeur 3D Animation
try Cascadeur
Once Ai can pilot software it's over for learning curve and will be unstoppable
THIS IS AMAZING! will try it soon, for now thank you so much!! u cool
Thanks so much for this video. One question. Can this process be used for also creating videos with multiple custom characters?
I think you live on the future! Amazing work flow as always, all tied up in an interesting and fun project.
Haha If I did it would be a one click and done 😅🤣 Thanks so much 💕
@@promptmuse haha very true
that is an intense tutorial. thanks for sharing!
No trouble ! Thank you!
I am just learning and this is exactly what I would like to create, however I am unfamiliar with each software and process used. Is there more breakdown and detail into each one I can follow? Or take a course or tutorial on each step? Thank you so much.. I’m in the hyper-focus train and can’t get off! Great videos ❤
I use Charcter Creator and UE5 to set the scene make a screenshot from the scene after i am happy with it and then feed this thru SD & ControllNet and have it illustrate. Works quite good to get what i want.
Love CC. That is a brilliant workflow you got going on.
Great tuto thank you ^^!
Your channel is a god send! Absolutely stacked with information and I'm so happy I found it!!!!
p.s. Davinci Resolve is totally worth it. The deflicker and automatic dirt removal tools are S tier but I'm a little bias lol
I'm really looking forward to trying it out. I have to buy alto of software/ subscriptions so gotta budget so carefully. Hopefully I can get it at some point!
Agree !
Wowww nice dear thanks for this tutorial
i think prrompt muse is just perfect ! ❤ ❤ ❤ such a beauty
ur workflow videos help me, also ur gorgeous with that wig 😄
Wooow, amazing workflow. Not for the novice, some of those workflow apps need some heavy experience, especially the Reallusion ones. Auto1111 can make you kill people as a novice but the learning curve isnt that huge.
Excellent video! Congratulations.
Thank you so much ✨
You're totally cracking me up with the wig 😂
I run locally (Macbook pro slow) and thinking about pods, so thanks for previewing what I am in for 😀(looks fast)
No problem ! I’ve been adding it in as everyone keeps asking my GPU. It’s a lot cheaper than buying a new PC 😂 I asked Runpod if they could support my community in the last video and they gave me a bunch of $25 to give out with my newsletter. So respect to them for doing that 👍
Oh also the Fast stable diffusion template within Runpod was created by the Last for Ben developer and is the best one to use (I just have my legacy pod running so will switch the template at some point)
lol the wig got me by surprise
great video, thanks!
It felt comforting, so kept it on 🤣😅
Thankyou! love your hair too
The animation is so clear and smooth, so awesome! Thanks for sharing this!
BTW I recently started to use Civitai and people need to be aware that you might see things you didn't expect. I keep the site in safe mode, but some of the models from Civitai are trained on other models and you don't always know what those models are, so I always add the words "nude" and "NSFW" to my negative prompt, because you can suddenly get surprised by images you really didn't want to see if you don't do that. I wish they had a better way of splitting the models up because safe mode doesn't always work, but I guess it isn't that easy.
This is great advice ! Thank you. My CFG & denoise are never particular high so never really get NSFW (not safe for work- for those wondering aka nudey !) 👍
yes. that is why teaching your own style to base model better approach.
I think I love you. I'm watching your videos just to see your face and hear you speak. I dabble in Blender a little but this is way over my pay grade. Cool stuff. 😍🤪
Hey muse, one thought. I'm also using the cc4/iclone pipeline in more or less the same way.
I'm finding that I'm having the best results with the Truebones mocap library. It's this huge set I discovered a while back when I was talking to people who make virtual influencers. It's just better than the reallusion actorcore mocap. Smoother, more lifelike, and it's massive. Just my two cents. Love ya
Awesome, thanks for that. I’ll check that library out. I’ve not come across that before so thanks for the heads up. You two cents are worth a million. Thanks again 🫶
Thanks! Do you know what it's for control m2m sd ??
Should you see activity in the Terminal once you run the script?
For some reason stable diffusion keeps freezing when I try to do multiple IMG2IMG on my end.
Not sure if the script is running or Stable difussion is crashing
You got me with the Scarlett model
Anyways, I think this is amazing. Now let's see what we can manage with it.
I just bought yesterday an RTX 3070TI to start building a PC for stable diffusion because of Xanthius's animation, this is crazy!
That is awesome! He is awesome !! Have fun. He only made this script in 3 days… Imagine what he can do with more time. He’s a legend !
@@promptmuse Yeah! I was that enthusiastic when I saw the results ^^
@@joannot6706 😂🤣 Tbf it’s a good choice !! I’m gonna get a Pc , in fact nerdy rodent another youtuber made a nice spec for me. It’s got to have both pc and cloud I think 👍
Amazing video!
Hey, thank you so much for your video ! Do we need a Lora to get consistent faces ? I tried everything but I don't manage to get a consistent face
Hey just doing a video for this :) Snowing how you can train a face with Lora
awesome video again as always. Can I ask how long did it take stable diffusion to complete all frames and how many frames did you have?
It was only 512x512 so only a few minutes. It’s says an hour on there as I went away… chilled and came back 🤣
Heya! Just putting on Frame Blending won't do anything. Pop on Time Remapping and move the end or start keyframe to offset the frames and force it. Remember to also set it to use Pixel Motion. That'll help a TON. I'm using the RE:Vision deflicker and the results are great using the remap trick. Gonna try out the Mult outputs now as well :)
I’ll give that a go, thankyou ! There are ppl like you that are pro at AE, I could learn a thing or two from 👍
@@promptmuse Oh and before I forget. on the Mult script, your Loopback source should be "Previous Frame" :)
@@kaelside hey yeah in the docs but found input worked better !
@@promptmuse Awesome! I'm gonna experiment :)
@@promptmuse I have experimented and “Input” has been more consistent, but I don’t get the correct style on occasion. Thanks for the tip 👍🏽👍🏽
girl you look like the character ;), congrats and thanks for the tip
Thanks you’re awesome!
thanks
Thank you! This is amazing!!!
Oh wow !!😮 Thank YOU do much 🫶
@@promptmuse ❤
Thank you.
This is incredible content
Wow.Super Beitrag.Alles Gute.
awesome as always! And a big yes please to a 'creating lora files' tutorial!
Yay okay… ON IT!
Does it work when the character is far away? like running towards the screen? i built my own model and it seems to only work when its close up and your character is close up also.
Yay! Great Job!
Thank you! Cheers!
That is so awesome
Great tutorial! What should I do if my gpu runs out of memory when it runs the control net image sequence script? I have an rtx 3070 which has 8GB of VRAM, so I feel like it should do okay, but for some reason I can only get through 2 images before it says I’ve run out of VRAM. I tried checking the low VRAM box on control net too, but that didn’t help
Yeah will be your video ram sadly. Tbh it usually does not take up too much. The are other programs such as run diffusion that you can use (but quite a bit more pricey than runpod) :)
How nice to see that someone else has Control Panel pinned to their Taskbar 😁
😁
This is amazing! Wow
Why, thank you 👍✨
Esta mujer parece hecha por IA ❤
Thanks!
Wow!!! Thank you so much !!! I really appreciate that 🫶✨