New AI Video Models & Big Updates from Hailuo, Runway, Adobe & Krea!

[@FilmSpook]

That short film someone made on MiniMax using the new character reference feature is TOP NOTCH and is blowing up. I luv it.

[@ShawnFumo]

Any more info on that to search for it?

[@rdthrthth]

which short film?

[@ManoOne-Music-Production]

Title?

[@mysticmaze2268]

Very Nice ... the 3D stuff is ALMOST ready to be useful.

[@TheoreticallyMedia]

I think we’ll see it this year. I’m with you, it’s SO close, but still not quite fully baked. Few more months in the oven I thinj

[@thisisreallychris]

The quips are really starting to get to me. My gah I love them so much.

[@TheoreticallyMedia]

Haha, the jokes are like water torture. Drip, drip, drip...but eventually-- I'll make you laugh!

[@PhilAndersonOutside]

It's nice to see continuous progress.

[@TheoreticallyMedia]

One foot in front of the other! We’ve come MILES since this all started! It’s awesome to watch in real time! (Although I sometimes wish we could FF a bit! Haha)

[@WhatTheFrenchToast]

Actually Tim looking at the diffusion I couldn’t help but think…. It’s Wonka vision! 😂

[@TheoreticallyMedia]

Haha. You’re right, it IS Wonka Vision!

[@AiObinnaLaunch]

Ai is getting crazy. Best time to be as alive. Great video man, subscribed

[@TheoreticallyMedia]

Appreciate it!! And yeah, such an insane time! Love waking up every morning to see how some breakthrough or research just dropped. You can see the highway to the future being paved right now

[@AiObinnaLaunch]

@TheoreticallyMedia mehn it's scary atimes for real fr. Ai is growing faster than human evolution lol

[@ge2719]

8:10 5 years ago and that would be on par with what you'd get from a full expensive camera system setup. ad its just from still frames in a video. thats crazy decent output. could even be usable for say holograms in a video game.

[@jamesyoungerdds7901]

Tim, long time fan and thanks for all the great content. I watch all my YT vids at 2x but yours is the only one I slow down just to catch all your charm in delivery. Really great stuff, thanks again!😀

[@phu320]

he does add a funny sense of humor to almost every comment, which I also enjoy

[@krectus]

Facelift looks really good, already working with video inputs too, nice. Hopefully they can fix the jitter cause that's impressive. Hopefully it can work with more than just people, because other single image to 3d options aren't nearly as impressive right now.

[@TheoreticallyMedia]

I’m sure that jitter will be solved before we see it eventually released. Although, off the top of my head, I was already thinking you could probably take that output through a Video to Video pipeline and knock it out.

[@krectus]

@@TheoreticallyMedia yeah, seems like it's very fixable. Wish they would release it just with image only for now, would love to try it out.

[@electroncommerce]

You always bring the goods!!

[@paullenoue8173]

Where can I find that crazy video at 1:22?

[@TheoreticallyMedia]

Haha, I think that's my favorite AI Video of all time. It was made by one of the mods on the AI Video Subreddit. If you go to Reddit and search "ai video avalanche" you should find it. It's a work of art.

[@os5037]

I am on the unlimited runway plan, can’t see the 4k upscale option weird

[@romanwhite2301]

Same. I don't see it.

[@TheoreticallyMedia]

Runway always has this slow rollout to users. I might have just gotten lucky. But, the announcement is out, so you should have it now? Or at least by the end of the day.

[@StopWork-ai]

His name. is Tim. Masterfully done.

[@TheoreticallyMedia]

Appreciate it!!!

[@julian42x]

does the hunyuan model let you upload your own videos to get sound generated for them or only their own videos?

[@TheoreticallyMedia]

I’m not sure if they have that on platform, but the model itself is MMAudio, and it is open source. If you use Pinokio, I believe you can download it there and use it locally!

[@aivideos322]

Look into latentsync. Its audio to video lipsyncs.

[@TheoreticallyMedia]

I’ll check it out! Thank you!

[@jaysonp9426]

4:33 actually made me lol

[@gumvue.studio]

I try Minimax image ref. , works like a charm, , waiting next the lypsinc. then we are done. I will upload the video soon

[@AIForWinner]

A major highlight is the introduction of Das or diffusion as Shader 3D aware video diffusion, which represents a breakthrough in video generation capabilities. Thank for your sharing

[@Dr.UldenWascht]

"TransPixar ... The Pixar movie that is more than meets the eye." You cheeky Tim you 😄

[@TheoreticallyMedia]

Haha, you were there in the 80s I see!!

[@inkandgrowrich9793]

Is adobe's Face Lift just for faces? or could it create a 3d version of a whole body... like say a ballerina twirling or martial artist doing a jump kick?

[@TheoreticallyMedia]

Currently it looks like just faces, but I’m sure they’re working on full body as we speak

[@EllenVaman]

Tx again matey 😎 btw what is the best way to get rid of watermarks on videos I generated with hailou ? Cheers 👍

[@TheoreticallyMedia]

I gotta look around for a watermark remover, but they always come out smudgy for me. Usually I just add letterbox bars. Haha. Now it’s “cinematic!” I think they’re supposed to be adding an option in soon

[@EllenVaman]

@@TheoreticallyMedia I agree there is no option to do the job well yet... I have tricks up my sleeve so ... all good ;)))))) Tx hun

[@bossgamer6583]

Illustrious one, I am in Mozambique and would like to know how to use Google VEO 2. I ask for your help.

[@Michf666]

Hailuo img to vid .. THE BEST .. cinematic colours and character movement

[@AG_before]

TransPixar - 🤣 That was great!

[@leandrewdixon3521]

Man, I just want to take a moment to appreciate Theoretically Media and Curious refuge. I follow everything in the broader A.I. culture and get so sick of the clickbait, hyperbole, sensationalism and blatant inauthenticity. The guys in the AI Filmmaking space, Theoretically Media and Curious refuge, are informative and engaging without all the nonsense. Much love!

[@TheoreticallyMedia]

Oh, thank you so much! The CR crew is great too-- Caleb and I keep threatening to do a podcast. Seriously, get the two of us in a room and we'll talk about this stuff for 3 hours. Not even kidding!

[@l-dixon4231]

@@TheoreticallyMedia Do it! 💪

[@Joyfulmischief]

Any hint on when adobe firefly video will go live ?

[@TheoreticallyMedia]

I'm actually talking to Adobe at the end of this month, so I'll see what I can dig up!

[@Joyfulmischief]

@ thanks ! It would be awesome to have it integrated to after effects 🔥

[@Maylin-Xiong]

Very nice ❤❤❤❤❤❤

[@BuioPestato]

Tim is The Man!

[@TheoreticallyMedia]

Bless you!! And you as well!!

[@dwainmorris7854]

DAS with flux looks interesting you can animate a character in AZ brush bring it over and have it skinned into a scene Maybe it will solve that consistency problem

[@Gibonsli]

Still no cartoon characters reference model...

[@TheoreticallyMedia]

Not yet, but I saw some promising stuff with Runway today. I’ll dig into it!

[@michail_777]

That's awesome! I was just wondering if they've added what they promised last year. I don't use Premiere Pro, so I'm not sure. I asked people who use Premiere and they told me that they only added a button:)

[@TheoreticallyMedia]

Yeah, I edit in Premiere, nothing yet. At least not in the timeline view that we saw last year. I think I’m supposed to chat with them soon, so I’ll see what’s up with that!

[@michail_777]

@@TheoreticallyMedia Thanks for the answer, I'll know.

[@harrygoldhagen2732]

Was that a reference to Independence Day at the start?😁🗽🚀

[@TheoreticallyMedia]

Haha, not entirely, but yeah- likely why it popped in my head! Someone in the comments was like “dive in” is boring, and I was like- you’re right! So I’m going to try to say “Let’s get started” in as many weird ways as I can! Haha

[@heartshinemusic]

The FaceLift thing is interesting for creating consistent characters, to get more input pictures of the different angles of your character, you can pick the ones that look most natural and convincing. It's a shame that it's from Adobe, lol. But I bet more will follow soon. Having a 3D model of your characters could be very helpful to create images that you can use as a starting point (first frame) for an A.I. video. (A 3D model would also be useful to dress your character with consistent clothes.)

[@vicesrus]

I tried out the character reference today but still seems to be some kinks in it, overall though it's working well

[@TheoreticallyMedia]

Yeah, agreed. I think the one shot aspect is a little limiting, but it certainly feels better than it did a few days ago. I’d say in a few weeks, it’ll be a killer.

[@insurancecasino5790]

Code does not work. Weird.

[@you-share]

Thanks!

[@ssotkow]

9:57 Only one generation is permitted for newly registered free accounts, a bit stingy. Thank you TM for the heads up nonetheless. Wishing you well sir.

[@TheoreticallyMedia]

Huh, I was told 6 early? Maybe they throttled it back? Eh, in a weird way I get it. The costs are super high, and Krea isn’t one of the massive AI platforms. I actually think it’s just like 4 guys. …but, really nice guys! Haha- maybe swing back over the weekend when there is less traffic and see if the rate improves. My guess is that it’s throttled depending on usage.

[@mik3lang3lo]

I think ALL of us prefer upscale that doesn't change anything and only upscale ❤

[@TheoreticallyMedia]

Haha, I don't know-- I'm a fan of those creative upscalers (SOMETIMES) to add a little zazz onto an image or video. But, yeah-- the problem is always in the consistency.

[@-Deena.]

🤔@5:28 I don't get the joke? 'More than meets the eye'. What's the joke? ❤

[@TheoreticallyMedia]

Haha. It’s an 80s kid thing. The original Transformers tv show had a theme song/tagline that was “Transformers: More Than Meets The Eye/Robots in Disguise” Dating myself for sure with that one!

[@-Deena.]

@TheoreticallyMedia Ahhhh! I am no Transformers fan, but I do love your channel ❤️

[@addsound]

Light the fires? Might want to think of a different phrase.

[@TheoreticallyMedia]

Yeaaaah, that didn’t age well. It wasn’t actually in reference to the fires, more on the Independence Day side, but yeah… You safe over there?

[@MabelYolanda-c9i]

2:13 What you’re explaining here is pretty much how any video model works….Luma was the first to implement this, but now all models follow the same architecture. So nothing new here.

[@TeddyLeppard]

DAS ist güt!

[@apatsa_basiteni]

Huh cool stuff.

[@Cyborg_VideoEditor]

like aways, im rooting for u..

[@TheoreticallyMedia]

Appreciate you!!

[@barryobrien1890]

Seems like we have hit a bit of a wall and most of these changes are just tweaks. Looking forward to the next big step as most of this stuff while impressive falls short of being useful

[@TheoreticallyMedia]

I get that. We’re in a build phase again. But I think a lot of those baby steps eventually culminate into a big jump. It might not be as “wow-y” as those first moments of seeing AI Video in 2022, but it has also progressed very quickly in just 2 years. Still, we don’t get to super useful without iterating and experimenting right now.

[@JayJay3D]

Hi Tim - I'm early :D

[@TheoreticallyMedia]

A wizard is never early nor late!! You arrived at just the right time!

[@JayJay3D]

@@TheoreticallyMedia 🙏🏻🙏🏻🙏🏻🙏🏻

[@ПётрПавловский-щ1х]

Transpixar :D

[@TheoreticallyMedia]

Can’t wait for the Michael Bay reboot of TransPixar!

[@virgilxavier1]

You know I couldn't put a finger on it for a while, but I just realized your folksy remind me of Dan Rather.

[@TheoreticallyMedia]

Haha. Never heard that one, but I did grow up with him, Koppel, and Brokaw in the background. Maybe something seeped it?

[@denemesoyadın]

I understand that you can't explain it because of sponsorship. I found sites that create "videos". The more I research, the more confused I get. My English is at an intermediate level. Video production minutes are not enough. I'm thinking of an unlimited plan and lip sync. Which one do you recommend to me? I just subscribed to your site. Your videos are in my area of ​​interest. I was afraid I would pay for the wrong subscription. Thank you in advance for your help.

[@HistoryViper]

Your hair looks good even on your bad hair days.

[@TheoreticallyMedia]

Haha-- if you listen closely, you can hear chainsaws in the background. The neighbors were hacking down a tree today. It was just a overall mess of a Thursday! My hair was pretty stressed out.

[@GrizzlyBossTV]

Yep next step is I think adding sound but everybody probably waits for more than 5-10 sec generations. 2023 was 2 sec. 2024 was 5-6 sec. 2025 is 12 sec then? Lol.🎉

[@TheoreticallyMedia]

By 2080 we’ll finally have that full hour we’ve been waiting for! Ha!

[@aivideos322]

Sound is almost there, audio to lip sync works, TTS audio works, making music works, its all there.

[@alexfilma16]

What rhymes with orange? No it doesn’t.

[@ПётрПавловский-щ1х]

I Set My Friends On Fire

[@TheoreticallyMedia]

Somewhere a Door Hinge is weeping.

[@morpheus2573]

Blancmange (Pronounced: Bla-monj) It'd pass in song lyrics

[@alexfilma16]

@@morpheus2573 The great Morpheus. We meet at last…

[@morpheus2573]

@ ...and you are?

[@elmentor8919]

I like that your videos are a good strong 10 minutes if it was 20 or 30 minutes I would’ve probably skip quicker, I got more things to do but 10 minutes of my time is OK.

[@parkerpierce8931]

you do know why people hate adobe right?

[@TheoreticallyMedia]

Oh, a few hundred reasons off the top of my head. And as someone who uses PS and Premiere, like daily- I get it. But, this work is from Adobe Research, which is kind of a whole other thing. We get mad at Google all the time, but I’ll never diss the DeepMind team.

[@Taytodude]

Cheeseburbner

[@TheoreticallyMedia]

Only if it’s as dry as a Applebee’s meatloaf

[@stereotyp9991]

They're all wasting time trying to accomplish AI video while no single entity has properly figured out images yet. Photos of people are still poor across the board. Recraft with its expensive creative upscale are the only ones somewhere close.

[@TheoreticallyMedia]

I think it’s all sort of happening in tandem. And that’s ok- we’re still in the infant stage of this technology. Wild swings all around, but every once in awhile, something hits and pushes the ball further than we expected. Will it be this 3d method? Time will tell, or maybe it’ll spur an idea for another technology. Y’know, I did a video on Recraft a month or so ago. Yeah, it’s pretty good!

[@thedevo01]

Ummm... That's totally wrong. Generations of humans have been figured out almost a year ago before Flux was even released. Edit: Yes, I mean perfect hands too. You just need to supplement the right files. Be that textual inversion, negative promoting, or LoRAs

[@Raulikien]

They need video for robots' world models, images are good enough. But images will get even better when they perfect video.

[@latentspace_fi]

Ummm. That does not seem to be true at all.

[@W-meme]

Photos of people are not poor at all. Flux can output photo realistic high quality realistic images that are indistinguishable from real life.

[@domehouse79]

Most of the time you’re showing examples and then you state that you know they aren’t very good examples. 😭

[@anarki9991]

Is this the big update? wow guess real artists don't have to worry too much. Love how this "expert" gets the 3D shader wrong. Just goes to show that when you have artificial intelligence you don't need real intelligence. But maybe you should read a book or two, in case the bubble pops. Take care.

[@MONTY-YTNOM]

To many now

[@raymond_luxury_yacht]

Ummm really annoying appropriate comment? I guess you're not CA? Or maybe it's ok cos we don't actually need Hollywood anymore anyway?

[@interestingt0pics]

very bad review

[@derekcunningham6456]

I am just waiting for the open source version of veo 2, but i think it might be a couple years.. >_

[@TheoreticallyMedia]

Haha, It'll likely happen, but I also think it'll be a ways off-- notably due to the training data. Not to say it can't/won't happen-- but by the time we get "Home Veo-2" the platform tech will be at Veo-10-- so, it'll look pretty outdated.

[@derekcunningham6456]

@TheoreticallyMedia yeah you are probably right due to googles willow chip. One can dream.