Create Amazing Videos With AI (Deforum Deep-Dive)

This is a deep-dive into Deforum, the AI animation tool that runs inside Automatic1111. In this session, we show off all of the cool features and how to generate amazing results with Deforum. We also discuss the benefits of doing it in the cloud vs. locally. Special guests join us on the call, including one of the original creators of Deforum. Everything you ever wanted to know about the past, present, and future of Deforum as well as step-by-step guides on how to use it are covered in this comprehensive livestream. Enjoy! :)

Resources Mentioned:

Timestamps:
0:00 Starting
1:20 introducing guest (revolved)
4:20 intro about what deforum is
11:50 overview about run-diffusion
25:15 playing around and explaining Deforum options part1
51:05 giveaway break
54:15 playing around and explaining Deforum options part2
1:03:20 introducing the second guest (huemin)
1:05:05 explaining diffusion module
1:07:20 Q&A where Deforum name came from?
1:08:50 diving into 3D
1:14:00 Q&A will there be any Adobe AfterEffects extensions for Deforum in the future
1:19:48 more insight about AI progression and Deforum options
1:24:35 explaining "Hybrid Video" section
1:37:03 Q&A Deforum for commercial use
1:39:41 explaining more settings for the output
1:43:25 Giveaway
1:46:30 Cool things you can do using settings (Music, Parseq, other tools)
1:54:16 Q&A explaining what Seed is
1:58:10 Q&A will Deforum will be available in other languages?
2:00:18 explaining Leaks and merging
2:04:50 more Q&A and giveaway
2:21:00 where is AI video generation going?
2:29:00 MRI stuff
2:32:16 more giveaway
2:34:50 Wrapping up

#deforum #aivideo #aiart

[nicolashuve]

Just want to say how much I appreciate this community. Film guy here with a motion design project to deliver in three weeks. I told myself "why not try deforum?", And I've just spent my Sunday listening to this stream and testing a few stuff. Now I'll definitely go that route for this project and it's gonna be dope. Sky's the limit.

[reallybigname]

Thanks for the shout-out! Love seeing a Deforum UI stream.
I just have one correction... Nothing that I have done is derivative of Sxela's work. While the work is similar, I've never opened his notebook. I came up with the RANSAC stuff after initially trying to mix video in and realizing that the render didn't move with the video if it was being mixed at a lower alpha than 1. After figuring out Perspective and Affine RANSAC modes, I also found the optical flow stuff in the cv2 docs, and incorporated some optical flow styles as well. I only knew that optical flow had been added to DISCO Diffusion in the past, and it was obvious that Sxela was using flow as well, just from his videos.

Some are surprised that I've never tried WarpFusion, but I do it on purpose. If I wasn't a developer, I'd probably support his Patreon and use his thing. I don't use it on purpose, because I don't want us to do the same thing as each other. I could have just analyzed and copied code from disco and from his stuff, if I had wanted that. I did not. I think it's better for us to be innovating in different directions.

I didn't stop with hybrid video either. I did normal animations before I did hybrid video! So, I recently incorporated optical flow into normal cadence for 3D mode, fixed the long-standing 3D depth stretching bug, and came up with some other redo features for rendering. I hope I can continue to innovate at the pace that I have been!

[DanielSvetlof]

Thank you for your contribution guys. Great info with such a good mood! Would love to see more of these live streams if possible.🔥

[marnusflatz]

Hello from Switzerland 😃…great having your guys live😁💪🏻💫

[jason-skoi]

Thank you both for the time and valuable information! Signed up for Run Diffusion.

[LonelySenpai]

Thank u so much for what you've done with this channel. you honestly blow every other ai art teaching content creators out of the water. & I'm incredibly impressed with RunDiffusion. It's incredibly inexpensive for what they offer! Feelin blessed to have found ur channel recently

[DrNothing]

Thank you so much for making this video!
I wish I'da caught it live, but am glad to have it now that I'm generating vids with deforum.
new stuff to learn and play with.
Thanks again, Matt! :)

[vidokk]

thank you very much, can you please enable automatic text (cc) for us users that simultanesly read what you speak, because english is not out first language, and topic is very demanding. thanx for the great content, nice guitars 😊.

[JulianHarris]

Very useful topic. Next time if you could make the video resolution at least 1080 so we can read the text that’d be golden 🙏

[GhostRat__]

This machine spirit is strong! May it prosper and bring us to new horizons

[GuyTheAnimated]

man I love this Ai stuff. I just ordered myself a 4070 just because of my absolute passion for the future of Stable diffusion

[sailcat]

You guys rock! I just like that you all are also artists who are embracing this amazing technology. It's funny that people say that artists are luddites but throughout history artists have ALWAYS looked to new media to express themselves and humanities ever-changing relationship with the world around them.

Keep doing what you're a doing and I'll keep thumbs upping your work!

[yourbartender]

I am on a mac pro laptop, and have been happily running sd, a1111, deforum with extra settings for a while. I know there are issues, but I can explore around them. There are also some out-of the box apps which run on a mac, including core-ML optimised image generation via mochi Diffusion (which is fast!) But I think one needs at least a m1…But yes, it works :)

[dineshkannaa]

Amazing session Matt 🔥Missed the live

[Alexhellbach]

you are killing it brother, keep it coming

[mischamim]

Great talk. I learned a lot and it's important to me now, because I started doing Deforum. I believe that it is in Deforum that at least animated films will be created. Deforum already has a lot that is needed to create animation and videos. It remains somehow to smooth everything out and stabilize. Much depends on the models, but a lot also depends on the control functions. The only thing I would ask more next time: what video, images, frames, depth map masks and other things where to insert. These "paths to insert" video, video masks, images and more secondary masks are not a lot.But it is not clear what it affects. For example with the generation of 3D at all misses what it does. The beginning is there, and then...
It's good that you asked about the formulas at the end, or I wouldn't have understood what and why:))) And it's great that you then added by asking if these formulas can be inserted everywhere or not. Thank you so much for that.Thanks for the link to rentry org. This document describes what I asked above. But better to see and hear once than read 100 times:)) I do not know, maybe many people and me including lazy, do not want to think:(

[Andonchoto]

Using free google collabs is easy and great, paying for Deforum tools will come in handy if one already knows how to use it and needs it for a paid project. Otherwise, learn it for free and master it well.

[huntinggamer]

Great content brother we need everything your talking about so many people don’t have a clue

[GameDevNerd]

Love these videos, they have been inspiring for me and my efforts and bringing together a new AI/cloud software company. Thanks for sharing all the amazing videos and tips that help keep me up to speed! 🙂

[davidlankester]

AI currently: tell me you want to try DMT without trying DMT.