New FREE AI Video Generator & Feature Length AI Films!

We've got a new FREE AI Video generator, and the quality and results are super impressive, a few approaching Sora level of quality. Although, this does come with a catch, namely the generation length leaves a bit to desire.
That said, I do have a trick to share with you to extend out your generations.
We're also going taking a look at the latest research on Feature Length AI Movies. We're looking at MovieLLM, a research paper that will show us exactly how this will work.
Lastly we're rounding out with some drama at Midjourney- namely, they've got beef with Stablity.AI-- we'll take a dive into the issue to see what the deal is!
LINKS:
Haiper.AI: haiper.ai/
MovieLLM Paper: huggingface.co/papers/2403.01422
Video on LTX Studio: • New AI Video That Does...
CHAPTERS
0:00 - Intro
0:33 - Haiper Overview
1:38 - Haiper.AI UI Overview
2:36 - Haiper Text To Video
3:46 - Haiper Image To Video
4:46 - Haiper Example Videos
6:50 - Haiper Hack to Extend Video
7:55 - Feature Length AI Films with MovieLLM
8:39 - MovieLLM Workpath
9:17 - The Key to MovieLLM
10:34 - How MovieLLM is Trained
11:26 - The Midjourney Drama

[@wiseowl09]

Tim always keeping us up to date on all these crazy number of tools coming out - much appreciated. When my YT channel is monetized, I'll come back and pay you back :)

[@TheoreticallyMedia]

Oh man. Thank you! And never stress on that front! I know the grind of climbing that mountain of 4K/1k! Onward!!

[@ElectroBlep]

Thanks for sharing! I clicked the link and started using Haiper before the video finished. The 2 second limit is not great. So far nothing I've put in was even moderately what I asked for. Of course 2 seconds is not much time for things to happen, but I'm not having much luck at all so far. It's also very slow to render. And so far, everything looks like it's in slow motion. So extending them with retiming isn't going to cut it unless you want almost no movement, but in super slow motion. Excited to see where it goes, but I don't think it's usable for anything I want to do at this point.

[@TheoreticallyMedia]

Super slow in generating right now, but I also think their servers are likely getting slagged right now. The Hug of Death-- haha! I think it'll ease down in a few days. The 2 second generations...yeah, I think those are best served for using as establishing shots, or non-specific actions. Like, "A walk on the beach" or "Robot in a City" There just isn't enough time in 2 seconds...although, if you use the slow-down technique I mentioned, you can get some cool moody stuff out of it. The real win here is the 4sec SD stuff. This is basically Free Pika.

[@ElectroBlep]

Thanks for the response. I've been playing with it off and on for a few hours, and some results are better than what I had gotten when I wrote my last comment. For even the least important stuff I want to do, I really need 10 or more likely 15 seconds, so I still hate the 2 second limit, but I can see potential in this. I think I'll post a video with multiple regenerations to give people an idea of what to expect. It definitely doesn't seem to follow a lot of my directions, but some results have been interesting.@@TheoreticallyMedia

[@TeddyLeppard]

I think these models will be vastly improved once systems are able to ingest both scripts and final movies and shows to compare and contrast choices made after script went into production.

[@TheoreticallyMedia]

Agreed. That's basically what the MovieNet dataset is indicating in the video. I think it's fairly new, so it won't be long...

[@josuebenvindo]

So, in the future, only writers and directors will have job in the film business?

[@TanChoonHong]

@@josuebenvindo I believe all roles will enjoy higher productivity if they embrace this tool in their workflow. Innovate or evaporate.

[@ElusiveTruth]

"Only psychopaths use light mode" 😂

[@TheoreticallyMedia]

It’s like staring into the sun for a full work day!

[@politanene]

@@TheoreticallyMedia Shocked me this statement, because my visual problems is way easier to work, with the light on :)... I think was mean from you say this... just not very kind.

[@tomurbanek]

@politanene Totally agree‼️ Wtf?

[@StarvinitySaga]

Lol 😆

[@StarvinitySaga]

​@@TheoreticallyMedia 👍 👍

[@suzevidz]

Thank you for telling us about Haiper. I've made and uploaded a couple of music videos from some other information you gave us in the past. I'm having a little bit of trouble with Pika going blurry on me all the time and now Runway keeps getting brighter and brighter until it whites out. 😂 I'd like to try Haiper because I'm working on two more music videos now. Thank you, again! 😁🎨🎬🎶

[@TheoreticallyMedia]

Try Haiper on the Text to Video mode, and the SD version. I know what you're talking about, and while it does have some color issues, its nowhere near at the level of the others!

[@suzevidz]

​@@TheoreticallyMediaThank you! 😊

[@asciikat2571]

Love your channel man, thanks so much for your incredible content.

[@TheoreticallyMedia]

Thank you so much! Really appreciate that!

[@asciikat2571]

@TheoreticallyMedia ✌️👽

[@piggywahwah]

Hope you read this thru all the comments... I was banging my head against the wall with low par sound fx. Someone suggested myedit ai. The image stuff is pretty standard but the sound fx are off the chart. I'm just completing a mini movie and the sound fx were good enough to go in. Really glad I got suggested it.

[@TheoreticallyMedia]

I’ll check it out! Lots of interesting breakthroughs on that front. Have you tried Pika’s new SFX? I’m testing that today, but I’ll look into Myedit as well!

[@piggywahwah]

@@TheoreticallyMedia and 11 labs has just dropped. Didn't know of the other thanks. The meta one is absolute useless. Zuckerberg needs to stick to destroying teens self esteem with unnatural beauty filters. 🤣

[@giri.goyo_yt]

Mouth dropped open over your Quay Bros style clips! Excited to try out Haiper! Also, lol'd at your white mode joke. Keep em comin.

[@TheoreticallyMedia]

Love the Quay Brothers! And yeah, my mouth dropped at them as well! Both the HD and SD versions look so cool! I think it’s a perfect example of what text to video is so good at right now. I know everyone wants to prompt, like The Fast and The Furious or something that looks like it is a blockbuster movie- but, really, we’re still in the stage of weird art movies. …and I’m ok with that! Lean into the weird!

[@champagnebulge1]

That stop motion doll vid you made recalls that Tool music vid from the 90s

[@Shanghaimartin]

Love ya man, but I think you're insane calling that highway shot close to Sora quality 😆

[@TheoreticallyMedia]

Haha-- well, ok-- maybe the eye shot? But also, WAY shorter!

[@jamesportrais3946]

Is Sora any more difficult to use than the likes of Haiper, Movie LLM etc?

[@BuddyBearCreator]

btw - the data "from" those films are how this engine is trained so yeah, the images, dialogue, camera, lighting, production design etc is scraped and ingested into their engine

[@PerChristianFrankplads]

Great and informative videos, but as a writer I implore you to use a thesaurus to find and use more synonyms. It's one thing to say "really really good" once or twice, but if you say it for everything you're impressed by, the viewer gets bored by the repetetiveness. Just FYI! Keep it up, and you just won yourself a new subscriber anyway. :)

[@TheoreticallyMedia]

Haha. Yeah, you caught one of my verbal tics. It’s funny, I’m actually totally aware of “really really good”- I usually catch it in them in the edit and roll my eyes at myself! Thanks for calling it out! I’ll put sticky note of some alternate words on my tripod to keep them front of mind!

[@PerChristianFrankplads]

​@@TheoreticallyMedia: I guess we all should, right? :D Thanks for your informative videos in any case! I'm now following MattVidPro, Matt Wolfe, Two Minute Papers, a few others I can't remember, and you. I should be pretty updated on everything, I think...

[@twilightfilms9436]

In the industry we go by certain rules that I’m surprised no KZbinrs mention, when they talk about AI feature length films. Those rules are the “Academic Standards” a.k.a. ‘Smpte Standards”. These end user tools are light years away from reaching QC standards (like Topaz, as a brief example)…..and for those of you who are not familiar with these standards, in short, no movie theater nor platform will showcase your film if you can’t pass a QC…..HBO has the highest standard of all, and the reason sometimes you wonder why certain movie is not on HBO is only because of Quality Control matters…..

[@TheoreticallyMedia]

oh, totally. I can say with near certainty that no other KZbinr in this weird AI space has ever had to address notes in a Technicolor or Deluxe QC report. I have...a few hundred times. (That said, you just quickly learn to check the box that says "Artistic Intent" on all the issues!) The larger point that I've been making is: None of this needs to be Industry Standard. It'll eventually get there, but right now I consider AI to be...(I'm dating myself here) like CoSa. After Effects before After Effects existed. None of that was Industry or Broadcast standard. It was a tool that existed before the term Motion Graphics existed. I doubt even Adobe realized what it would eventually become when they acquired it. My overall point being: None of this has to conform to any standards. For now, at least. It'll get there, I'm sure-- but in the meantime, I think it is such a great time to bash these tools together in totally unexpected ways to see what comes out.

[@ivydune4185]

I keep saying: it's way too soon for AI to charge for their services. It's far from good and there's always something strange happening, and I don't mean AI hallucinations.

[@TheoreticallyMedia]

Given the cost of compute, I do understand the need to charge-- the way I see it, most of us are early adopters into this technology. And yeah, in some ways, we're paying to beta test-- but I also like looking at it as: we're shaping the future of this technology. ...I mean, for a monthly fee. That said, I've been giving Midjourney money every month since the beginning of v3. So, I'm just onboard for the whole ride.

[@ivydune4185]

Yes and no. I understand that people don't work for free and these things cost time and money. But they are working for profit. You're paying at least $120 (yearly) for a faulty product that you can use for limited amount of times a month (credits). Even if you have lots of credits, it's still 120/year per a limited usage with random and erratic results. If you buy a software, you pay that or less once in a lifetime and you use it as many times as you want per month with reliable results.

[@noNumber2Sherlock]

Tim, as always you are bar none THE authority. Thank you! Quick question, is there an AI generator yet that can take an image, lets say a person standing or sitting, building, etc, and then change the camera perspective to go up ; looking down below towards the subject?

[@TheoreticallyMedia]

Ah thank you so much! So, I know exactly what you're talking about and...although there are some ways to do it, it isn't quite what you're looking for. Midjourney is talking about this very thing for a future release. Currently, the only real "easy" way to do it is via 3d software with something like Fondant (I went over that here: kzbin.info/www/bejne/op-cc414aNx3q9k)-- after that, you're getting into the world of Gaussian Splats, which are cool, but require multiple angles to begin with-- kind of like the old Matrix "Bullet Time" shot. I'll keep brainstorming for you though! In the meantime, it IS coming!

[@noNumber2Sherlock]

@@TheoreticallyMediaTim, thank you. Nothing but LOVE man! Thank you, YOU make a difference!

[@PandaKattPk]

could you make the video longer by a screen shot at the end and generating it again?

[@TheoreticallyMedia]

Maybe! That's an old trick-- I'll give it a shot! It might just work!

[@K-10-limitless]

I've been so happy with Pika and almost forgot that we are in the very beginning of Ai so of course there will be worth contenders right around the corner. And this one is here. Cool. Another great share.

[@TheoreticallyMedia]

Sorry, missed it earlier! Yup, Pika is great as well! And now they have Sound Effects in the outputs! It’s crazy how quickly everything is happening!

[@jamesportrais3946]

Chaps, let's say I'm really good at one element, and @K-10-Limitless is brilliant beyond belief at another and we decide to pool our resources. Could an encompassing application remunerate us in accordance to our efforts? If so, I have just invented the worlds first meritacratic media resourse.

[@gumvue]

yes, I am testing righ now.. so the HD version is very cool.. image or text to video are good

[@TheoreticallyMedia]

Really looking through it, I’m close to calling it Sora level. Just…really short. Really short! Haha

[@gumvue]

yes, so far , am now rendering all my images for a new video I try to make last frame trick to make it 4 sec, but it lose the color .. so need fix colors and sharpnes before upload.. maybe if go and fix with fofooocus and make 3 4 2 sec videos , then can get 6 sec ones, HD level. @@TheoreticallyMedia

[@Nexusnigma]

is it for free?

[@PhilAndersonOutside]

The issue I have with Movie LLM is that while I can definitely see a day soon enough where this kind of model will be heavily used, I don't think it's anything close to what a true creative power user will be interested in. I could write a treatise on this, suffice to say I believe creative AI "filmmakers" in the future are going to seek control over the script to storyboard, and scene by scene control and guidance capabilities. I've yet to see anyone begin to tackle this approach seriously, and to be candid, despite efforts to comprehend how to do it it's something bigger than my shrew sized brain to attempt to even begin to develop on my own.

[@TheoreticallyMedia]

I totally agree with you. I think this technology is better served in games or interactive media. Like, a writer/creator comes up with the “golden path” but, the viewer player has agency to explore the world. The LLM would allow for this, while gently nudging the viewer/player back on the road.

[@kevnar]

How long until we have full text-to-film capabilities, where you can type in the story you want and the AI creates the video, dialog, sound effects, and movies score? Five years? Ten?

[@TheoreticallyMedia]

oh, far less than that. I think we'll see something in the 20 minute range by the end of the year. It might be more suited to animation...but I think by end of next year? At this rate, we'll be there. I'm not sure it'll be commercially available-- as that is going to take a ton of compute power-- but it'll be here.

[@samthesomniator]

Me clicking the video: "Nice, how will AI help me as a video creator? 😊" The video: "...entire movies from one prompt" Weeeeell. 😑😂

[@TheoreticallyMedia]

Haha, to be fair, that’s still just a research paper. I have my doubts that it’ll end up catching on. I see this kind of tech eventually ending up more in games or (perhaps) interactive features in the future

[@melgspawsome]

I have a lot of problems with generators morphing and/or warping the final product. Eventually, they'll get it right, since a lot of these programs are in their beginning/testing stages.

[@TheoreticallyMedia]

Yeah, I’m sort of waiting to read that paper where they come up with a method for the model to separate the foreground and the background into two different attention areas. Right now, the model sees an image as one thing and it loses track in highly detailed images.

[@melgspawsome]

@@TheoreticallyMedia Yep, this is true. I started separating the background and characters into their own images. I found out when I generated a background into a video; without any characters, vehicles, etc in it, (i.e. a focal point), that it moves much smoother without warping/morphing. I discovered that the camera movement generated, when overlaying a still image of the character, makes it appear as if the character is walking. For example: I used the prompt "a noir city back alley late at night, cold, wet, rainy, no people, no vehicles" and it generated it smoothly (I used Leonardo Ai for that though), when I generated the still into movement, it looked like a first person view of someone walking. I've been having one heck of a time trying to get characters to generate them walking without morphing or warping so it seems that's working for now until Ai programs get it fixed.

[@Crazy_Truth]

Powerful computing is also necessary to accelerate the generation!

[@TheoreticallyMedia]

oh, a massive amount of compute. Did you hear that it takes Sora 1 hour for every minute of video? Could you imagine what that render time would be like if they released it?

[@Crazy_Truth]

@@TheoreticallyMedia 😲 why they don't use my dual core PC 🤔

[@funkymonkey529]

I've been looking into this for days now. Which one lets you make a video longer than two seconds??

[@TheoreticallyMedia]

Haiper is now up to 4 seconds, which is kind of the current standard, outside of Sora. But, obviously, Sora isn’t publicly released. I’ve seen a few poking at the 5 second range now. And, you can always do the Final Frame trick, where you take the last frame of your video, then use that as an image reference. But, the video quality does tend to suffer a bit.

[@funkymonkey529]

@TheoreticallyMedia How are people making these big elaborate videos? Everybody talks like they know what they're doing , but nobody actually posts one. I know some people can. Every time I use one of these AI More than once it'll never Keep what I've seen the first time and add to it. It just changes it all or doesn't change it at all. It's frustrating because to me it seems like AI is just a false claim and absolutely worthless.

[@Avalon19511]

When these other video generators go as long as Sora then I'll pay attention:)

[@TheoreticallyMedia]

I hear ya-- but, as I often mention: Sora isn't out yet. And to be totally honest: I don't think it will be for quite some time. I could be wrong, but I think they're going to have that locked up until much later this year. The 2 second thing is just the starting point-- to be honest, the 4 sec standard is sort of where it's at. It's basically like Free Pika right now.

[@Whiztlex]

Seeing the rate at which these models progress. I'm positively optimistic that we get something better than Sora already this year

[@Avalon19511]

@@Whiztlex I totally agree

[@rodbarker6598]

@@TheoreticallyMedia It's still too close to an animated GIF for me. I'll wait until they start breaking a minute; it will come ai video is in baby step mode at the moment.

[@RSProduxx]

@@rodbarker6598 well, at least something like 20 seconds, which will be enough for a scene from a certain perspective. If you can get enough consistency to make a sequence from scenes, that´ll be fine. But yeah, 2 or even 5 seconds are way too short.

[@jefferson6459]

I really want to see which side the a.I would choose in a Mike Tyson vs Mohammed Ali fight in a video generator

[@TheoreticallyMedia]

Haha, that would be epic! Should feed that into ChatGPT and tell it to access their stats, both in their prime, to see who would come out on top. Speaking of Mike, Why?! Just WHY is he getting back in the ring?

[@rubic99]

All my AI interface having light backgrounds. In this case I’m a psychopath. Called my GP already.

[@TheoreticallyMedia]

Come to the dark side! We take naps!

[@HumanAki]

Tim, you are a godsend and I will forever be singing your praises from here to Valhalla! 👍😎👍

[@TheoreticallyMedia]

I'll swig my cup of mead to that!

[@maxwellcoleshow]

You are my new favorite channel. Omg. Love these posts as a creative. 🚀🔥 Fantastic. Fascinating. Exciting. Mindgasmic.

[@TheoreticallyMedia]

oh, man! Thank you soooo much! Really appreciate that! I'll keep grinding!

[@butonphillie5114]

Thank you so much you made my day from terrible to bearable, sent you a super thanks ! ❤ love your personality

[@TheoreticallyMedia]

Seriously, that made MY day as well! But, super happy to hear my little video brought a little sun to your day! Trust me, I know those days as well, so I don't take it lightly! Weekend is almost here! Treat yourself to a scoop of ice cream!

[@futrstory]

These AI generators are getting crazy! thanks for the info.

[@WoodenCreationz]

Awesome Generator and great video!!! Ty!

[@TheoreticallyMedia]

You're welcome! And thank YOU!

[@ChristianIce]

Tried it. Impressive, but the "repaint" feature is a lottery and you almopst always lose. It needs more controls on "how much" the AI is going to hallucinate.

[@TheoreticallyMedia]

Agreed. Repaint isn’t really there yet, but they’re also super early into launch, so I’ll give them a lot of leeway here. Also, I didn’t really hit it in the video, but you can’t specifically aim for target areas to repaint. It’s basically the whole frame. That said, no complaints. It’s free, and they’re just getting started.

[@thewebstylist]

I was stoked for Haiper but 2 secs yeesh Just tested but quality looks great.

[@TheoreticallyMedia]

I'm thinking of it as a preview of what is to come. 2 seconds for now, but in a month or so, it'll be 4 or more. And actually, the SD version is pretty cool in straight T2V!

[@K-10-limitless]

peoples attention span is 2-5 seconds for a shot. you can slow it down in editing, its really not that bad, but when all of these sites make you pay for anything over 5seconds you might appretciate it more. and maybe not.

[@gu9838]

haiper is OK........but it is limited to 2 seconds and when i tried it took FOREEEEEEEEEEEVER to render so not really an optoin for me.

[@TheoreticallyMedia]

Yeah, it's basically "launch day"-- or at least they went wide on press today. So, I'm sure the servers are getting slagged. Give it a few days and it'll probably get better.

[@steeleforce13]

Already getting better. Just give it time

[@melgspawsome]

Yeah I had the same problem and it would stop at 50% and wouldn't download, but I'm sure they'll work all the kinks out. I'll keep working with it and see what happens 😊

[@Paintingtube]

Thank you for showcasing your AI product on KZbin. However, it’s essential to clarify that while offering it for free is generous, it’s crucial to ensure that users understand the limitations, particularly regarding commercial use and the time it takes to generate content. Additionally, it's important to note that the content generated may not be owned by the creator, as it falls under the platform's terms and conditions for future purposes.

[@circelink]

Nicely reported Tim!

[@TheoreticallyMedia]

Thank you so much! Really enjoyed making this one! I think Haiper has a lot of cool stuff going for it!

[@TheDandonian]

I've been playing with it quite a bit today, using the image-video stuff, it's kind of ok, out of about 10 different rolls, I managed to get maybe a second or two of video that would have been usable. Not exactly ideal but i've seen a whole lot worse. I really hope they take off though. We need a lot of competition in this space, so that Sora doesn't end up owning the market and controlling the price. It's getting very expensive to use all of these great AI tools. The text to video stuff is very promising though. I gave it a prompt about Cersei from Game of Thrones standing in the red keep and the end result was something that looked 99.% like it could have been a shot from the show.

[@TheoreticallyMedia]

I think (currently) it's strong point is Text To Video. It does OK at I2V, but we still have all the usual problems there. Although, much more stable than say, Pika-- it still doesn't really follow directions. But, you can get some cool stuff out of T2V if you use the SD version and crank the motion up! The HD I'm treating as a little preview of what's to come.

[@TheDandonian]

@@TheoreticallyMedia Yeah I completely agree. For the longest time, I've wanted to create a specific scene that I think should have been in Game of Thrones and this is the first tool I've used where I think that with enough time, I could probably pull that off and it would actually look decent. If they manage to improve the Image-Video, so that we can extend videos and ensure characters are in the same clothes, etc.. It will be a big step.

[@spacekitt.n]

yeah it still looks like shit

[@BuddyBearCreator]

when you say "using" the data from the movie net data set - you mean they copied the data right ? they're training their model on copywritten material from creators. I think the lawsuites from the creative Guilds (WGA, DGA, SAG and of course the major Studios ) are ramping up now

[@NicoleSummer8]

Awesome and thank you!👍

[@TheoreticallyMedia]

1000%! And thank YOU as well!

[@butonphillie5114]

Thanks!

[@TheoreticallyMedia]

Oh man! Thank you SOOO Much! That really made my day!

[@mephistopolice]

I tried out Haiper, but it has some real noise issues. About .5 seconds in and all of the background characters look like they got snapped in a Marvel movie. Gonna keep toying with it, but it's not looking great.

[@TheoreticallyMedia]

I think the real win in Haiper (so far) is the T2V SD mode. You get a lot of useable results there, and at 4 seconds. It’s basically Free Pika.

[@nicheva417]

So basically you'll be able to watch every movie ever made but you can set it into whatever universe you prefer. Show me silence of the lambs but in the star wars universe. I think syncing the audio convincingly will end up being the barrier to entry for a while.

[@TheoreticallyMedia]

Hypothetically, that’s the idea (and MAN do I want to see that movie now!), but to be honest: I think this technology will be better served in games or interactive stories. Almost, “choose your own adventure” type entertainment, where a writer has the end point, but the viewer can still explore the world. I also think it’ll be awhile before the story/writing engine is hooked up effectively to the image/video generator. MovieLLM shows it can be done, but I figure it’ll be super buggy for quite some time.

[@TheBlessingReport]

Another great video as always

[@TheoreticallyMedia]

Thank you so much!! Seriously, really appreciate seeing you in the comments!

[@TheBlessingReport]

Yeah I like your comedy and style @@TheoreticallyMedia

[@Thefreequincy]

Thank you 🙏🏾

[@TheoreticallyMedia]

1000%!!

[@yummylife1]

Thank you

[@DIAT1234]

Hey I'm looking for a cartoon maker that allows you to use custom animal characters and easy to use. Any suggestions?? Create studio told me i can use my own characters, only to not be able to put action to them. 😡

[@CounterFlow64]

There are really good video interpolation AIs that are open source, best and fastest one I know is RIFE. Open source GUIs for it's use is for example Enhancr and REAL video enhancr (which is Linux only for the moment). Enhancr integrates loads of open source AIs and I would call it a competitor with Topaz. Better donate to these than buy Topaz imo.

[@tyberwhite]

I think it's important to note that "feature length" has never been an issue, but rather "feature quality".

[@TheoreticallyMedia]

Very, very good point! I’ll always prefer a 30 minute ripping short over a 3 hour mess!

[@SungLegendaryFigments]

You earned my like and follow with the psychopaths comment. 😝

[@paulroberts8695]

all these AI video generators are pointless when they only outpout a 2 sec clip, cant do hardly anything with that.

[@TheoreticallyMedia]

To be fair, Haiper is now at 4 seconds...which is kind of the current standard. It'll be 5 to 10 seconds soon enough.

[@kalish8437]

Only 2 on my app ​@@TheoreticallyMedia

[@matthewlarue1883]

Most movies change between clips that are 2-5 seconds usually. Can't you build on that?

[@kevinaustin0369]

I agree. Very misleading saying it’s feature length. Better off writing a book and using still images for it. Narrating throughout the film feels lame. Like they’re trying to force progress before it’s here. Not an IT guy, but if you can do 2 seconds what prevents from doing an actual feature length film? Processing power? Memory capacity? All the above? Looking forward to the day we are able to do so.

[@sughraanwar2979]

How to use for long videos

[@Isromancedead]

Part of me feels like waiting until 2025 when these softwares are really good, I feel all of them aren't great yet being new

[@tonyclif1]

So if i change to dark mode im no longer a psychopath?

[@TheoreticallyMedia]

Instant cure, from what I’ve heard! Just don’t rapidly flick it on and off like a light switch!

[@AmeenAI4.0]

👍👍

[@TheoreticallyMedia]

Thank you!

[@eri7-11]

I tired it haiper last night twice and it was stick at 50% for a long time then gave an error so I quit. I will try again

[@TheoreticallyMedia]

I think we might have given it the "Hug of Death"-- a lot of new users flooding in all at once. I think they'll ramp up quickly. I think by next week things should be pretty stable!

[@EllenVaman]

Funny and informative 😁👏🏻

[@TheoreticallyMedia]

Thank you! Always the goal! Just because it’s info, doesn’t mean it needs to be dry!

[@EllenVaman]

@@TheoreticallyMedia absolutely :) one question is there a free option to upscale short videos ? tx.

[@AzeAlter]

very very cool

[@TheoreticallyMedia]

Totally Agree! Free is always awesome!

[@patchadventures9247]

Has no one thought to generate a 2 second video and then grab the last shot of it and then put it into the image to video and just extend it that way?

[@TheoreticallyMedia]

Final frame.net can do that, and automate it! You just ended up with gradually lower quality video for every iteration. Until you end up with these blobs. Still, there is something to that technique and you can often buy a few critical seconds if you’ve got a real winner generation.

[@leekotvfilms5236]

Ai is evening the playing field….when ai reaches big budget movie quality. The most creative and best story tellers will be the most successful

[@TheoreticallyMedia]

Yup! It'll take more than ChatGPT writing a story for you-- you're still going to need to have something to say, and the story to say it with!

[@weaponizedtv5228]

Best comment 🎉

[@leekotvfilms5236]

@@weaponizedtv5228 🎥✊🏾i can see people using ai with augmented reality to make feature films…. It’s gonna be like a whole new realm of cinema

[@SUPERWARRIORMAN-xe8si]

Would be cool if you could just say, make Terminator 2 with Stallone, or make Aladdin in real life version

[@TheoreticallyMedia]

Haha, was that a Last Action Hero reference? Because I too, have always wanted to see that Stallone Terminator movie! Although, my eventual AI project is going to be the lost Bruce Lee Terminator film! I've joked about it enough on the channel that I've convinced myself I need to do it!

[@RSProduxx]

@@TheoreticallyMedia Bruce Lee vs. or AS the Terminator? ;)

[@LouisGedo]

👋 Looking forward to this video

[@TheoreticallyMedia]

Heya Louis! It's a pretty good one. Video gen is short, but I'm curious to see what you think of the quality!

[@LouisGedo]

@@TheoreticallyMedia Short is an understatement. I don't actually see much utility for such short video AI.....2 seconds??? Even 10 second clips have very limited utility IMO. I'm sure you don't need any more proof that I'm the real Debbie Downer of Video AI. The ONLY tool thus far that has me very excited about video AI is SORA......but mostly metely as a very good starting point for the utility of video AI.

[@LouisGedo]

@@TheoreticallyMedia In terms of quality..... it's eh? I do think that quality can't be meaningfully judged by just 2 seconds because an ESSENTIAL aspect of video quality is object permanence coherence consistency. That's why I made a big deal out of the time scale difference.

[@TheoreticallyMedia]

@@LouisGedo I’m thinking of the 2 sec generations more as a preview of what’s to come. Current use case is probably using my slowmo method to generate moody pieces. You aren’t going to get a ton of control, but I think it would be cool for stylish art videos. That said, 2 seconds will become 4, and 4 to 8 in no time. Still, not passing the Smith Test yet!

[@LouisGedo]

@@TheoreticallyMedia I'd bet that experts in this field will tell you that it's magnitudes more difficult to get 10 - 20 seconds of object permanence coherence consistency than 2 seconds everything else being equal. I'm definitely NOT an expert but it seems so intuitively obvious to me that this would be true......at least to a point. Maybe after a certain length, the difficulty in object permanence coherence consistency isn't such a major hurdle.

[@etopudoekene9083]

Only one second of the video I created showed up and no voice. It is not possible to ascertain its full potential.

[@TheoreticallyMedia]

Ah, for sure. To be clear: they aren’t doing voices yet. Pika just rolled out voices in video, but Haiper has not gotten there yet. The time limit is due to increase pretty soon as well, according to their Discord.

[@robertjunior1511]

can you monetize your youtube video if you are using haiper AI?

[@TheoreticallyMedia]

You can. Although, per KZbin's rules, it can't be low-effort. In other words, you'll have to put some work into it. Now, how do you quantify what that threshold is? I have no idea. Basically, if you generate a bunch of videos of people doing Yoga and put it over a 3 hour loop of Meditation Music, you'll probably get demonetized. If you make a short film? Probably not.

[@gRosh08]

Cool.

[@TheoreticallyMedia]

Thanks so much for dropping a comment!

[@Ghareebz]

guys does someone have a better alternative ?

[@agnesslovehealz]

Ahh my fave

[@TheoreticallyMedia]

Ahhh, bless you! So good to see you! Hoping you have a stellar weekend!

[@gunendra]

Fun thing bout ai is tht, makes ya do all the crazy things..

[@TheoreticallyMedia]

I love this period of tech. It’s early enough in where there are no established rules. This is like being a kid on a playground.

[@gunendra]

@@TheoreticallyMedia technology is fun no doubt about it, but it's dangerous, we were made to use our brains in some other way,the spiritual way,and God liked men to behave like little children not scientists , for eg,air doesn't have to be o2,n,c ,h etc.. it's been done before trying to make earth like paradise, we are also a bit like ai,bcoz we are also behaving like pre programmed bots,accepting the way of the world,we can only try to put it to good use,but it's negative impact is immense, it's all indulgence and waste of time,bcoz we r here for a short duration of life span,time eternity is our problem,the spirit is not dull energy,it's sentient,n that is another way of explaining, the soul ,algorithm,life gave life,matter gave matter,we are stuck in between...

[@gunendra]

@@TheoreticallyMedia @TheoreticallyMedia technology is fun no doubt about it, but it's dangerous, we were made to use our brains in some other way,the spiritual way,and God liked men to behave like little children not scientists , for eg,air doesn't have to be o2,n,c ,h etc.. it's been done before trying to make earth like paradise, we are also a bit like ai,bcoz we are also behaving like pre programmed bots,accepting the way of the world,we can only try to put it to good use,but it's negative impact is immense, it's all indulgence and waste of time,bcoz we r here for a short duration of life span,time eternity is our problem,the spirit is not dull energy,it's sentient,n that is another way of explaining, the soul ,algorithm,life gave life,matter gave matter,we are stuck in between...

[@gunendra]

@TheoreticallyMedia technology is fun no doubt about it, but it's dangerous, we were made to use our brains in some other way,the spiritual way,and God liked men to behave like little children not scientists , for eg,air doesn't have to be o2,n,c ,h etc.. it's been done before trying to make earth like paradise, we are also a bit like ai,bcoz we are also behaving like pre programmed bots,accepting the way of the world,we can only try to put it to good use,but it's negative impact is immense, it's all indulgence and waste of time,bcoz we r here for a short duration of life span,time eternity is our problem,the spirit is not dull energy,it's sentient,n that is another way of explaining, the soul ,algorithm,life gave life,matter gave matter,we are stuck in between...

[@gunendra]

@@TheoreticallyMedia @TheoreticallyMedia technology is fun no doubt about it, but it's dangerous, we were made to use our brains in some other way,the spiritual way,and God liked men to behave like little children not scientists , for eg,air doesn't have to be o2,n,c ,h etc.. it's been done before trying to make earth like paradise, we are also a bit like ai,bcoz we are also behaving like pre programmed bots,accepting the way of the world,we can only try to put it to good use,but it's negative impact is immense, it's all indulgence and waste of time,bcoz we r here for a short duration of life span,time eternity is our problem,the spirit is not dull energy,it's sentient,n that is another way of explaining, the soul ,algorithm,life gave life,matter gave matter,we are stuck in between...

[@MONTY-YTNOM]

Cube the original, was brilliant

[@Zoyous]

Agreed, it's best to disregard the sequels.

[@TheoreticallyMedia]

So good! And it still holds up!

[@TheoreticallyMedia]

I only remember seeing Cube 2...and really nothing else about it. Like, I know I saw it, I can't remember a thing about it!

[@MONTY-YTNOM]

@@TheoreticallyMediaI saw the trailer and just didn't want to watch it

[@FleXxTheFuture]

Does ANYTHING exist without registration??

[@TheoreticallyMedia]

These days. Barely. I mean, even my grocery store has my email address and phone number.

[@koen.mortier_fitchen]

Thanks for the video. It morphs like Pika and Runway. AI video is not usable yet imo.

[@TheoreticallyMedia]

Try the Text to Video side. Particularly on the SD model-- a lot less warp/morphing there! I need to check to see if we can neg prompt as well.

[@g0d182]

cool

[@anndavidson6950]

And up until now I thought only psychopaths used the light mode (what a way to learn something new about myself lol)

[@flagshipbowtie]

2 secs?! Why even release that?

[@TheoreticallyMedia]

Well, gotta start somewhere. I think Gen1 was like 3 seconds when it was first released. And again, I think if you speedramp it by 50%, you can get some cool stuff happening.

[@flagshipbowtie]

@@TheoreticallyMedia No it was 4 seconds free and 15 secs if you've paid. All you can have in 2 secs is just some shifting and basics happening.

[@TheoreticallyMedia]

@@flagshipbowtie On Beta launch? Back in the Discord days? You might be right about that...AI has rattled my brain at this point! I think the 2 second thing is useful for establishing shots if you can extend via slo-mo (as I did in the video) or catch a good loop. For now, I'm really thinking of HD as a preview of what is to come. The 4 second gen is where its at. We've basically got a free Pika here.

[@flagshipbowtie]

@@TheoreticallyMedia Pika is terrible. This one doesn't look much better. SORA fundamentally works in a different way that's why it looks like a real video.

[@LouisGedo]

Sorry to always be the Video AI Debbie Downer, but Haiper isn't even close to SORA. Let's say the average length of the videos Open AI gave us with SORA was between 10 and 20 seconds. That's 500 - 1000% greater object permanence coherence consistency than Haiper...... that seems significantly different to me. With a 1 minute video in SORA, that would be a 3000% difference. I warned you I'm the Debbie Downer of Video AI. 😉

[@TheoreticallyMedia]

Haha, I count on it at this point! I know things are going in the right direction when you chime in with an approval! It'll happen this year, I'm sure!

[@LouisGedo]

@@TheoreticallyMedia I'm 💯% in favor of any company or individuals working on video AI .......I'm just not easily impressed. The first several versions of MJ were crap as were the first 2 of Dall-E. Same with Stable Diffusion. But I was still 💯% supportive of the focus on and resources invested in image AI technology. With Dall-E 3 and maybe even MJ 5.1.......I was finally truly impressed with image AI.

[@VaikunthaStudio]

❤❤

[@TheoreticallyMedia]

Thank you!!

[@VaikunthaStudio]

@@TheoreticallyMedia Is there any AI tool for story board?

[@VaikunthaStudio]

​@@TheoreticallyMediaI did it but speed is very slow and there is a little blur in the image to video. And 2 seconds is a very short...

[@WisdomChronicles-el5ow]

Can i use in mobile phone and with free?

[@TheoreticallyMedia]

Haven't tried it on mobile, but if you're just using a web browser, I wouldn't see why it wouldn't work. Although it might not be optimized for mobile. But either way, yes-- it will still be free.

[@WisdomChronicles-el5ow]

@@TheoreticallyMedia thanks for reply me

[@phillipscott7653]

Hi good info..its on and off they must be working on it. but its been good and handy. thanks. i paid a fee. i find it good for the $

[@TheoreticallyMedia]

For Haiper? There is no fee. At least currently while they are in Beta.

[@phillipscott7653]

I paid a monthly fee to go up 👆 a level to premium, then a couple of days later the site was difficult to get into. They are doing things and charging their minds maybe. Maybe I was scammed. But I paid 6.99 for a premium month.

[@maxgreatcat7138]

"only psychopaths use light mode" - so true!😄

[@pinkpigsnack]

it was very good the only thing is only 2 seconds

[@JaxonBeckett]

Sora is now the new bar anything less will no longer impress at all (full stop)

[@TheoreticallyMedia]

Yeah, and that's unfortunate because at this point it could be vaporware. The problem is, we've SEEN Sora, but it might also be a glorified tech demo. Also, MKVHD recently did an interview with the Sora team where they straight up said: "It's not coming out any time soon"-- (kzbin.info/www/bejne/iaPLYp-rqZmEaKM) I know we've all seen the promised land, but there is still an ocean to get there.

[@weaponizedtv5228]

People speak about Sora as if it is available to the public. A lot of times when they roll it out it's not as good as advertised. It's not out yet because it's not ready.

[@JaxonBeckett]

@@TheoreticallyMedia CEO of Runway stated in a twitter reply that Gen 3 will equal and even be better than Sora examples what are your thoughts

[@jaysonp9426]

"Only psychopaths use light mode" truer words have never been spoken

[@TheoreticallyMedia]

Light mode is like staring into the sun for 8 hours (or more) a day! Who else but a psychopath would do that?!

[@jaysonp9426]

@@TheoreticallyMedia exactly! 😂

[@Content_Supermarket]

10:32 he is more like e'd sheren

[@TheoreticallyMedia]

Ha! Good call there!

[@Content_Supermarket]

@@TheoreticallyMedia 🥰

[@user-oq2bn4wq1y]

"I'd love to live forever, but just for a little while". -Cliff Spab

[@kidsbedtimestoriz]

is movie net free??

[@giggles7179]

The amount of people whose employment will become redundant because of what these generators can - and will be able to - do is stunning.

[@TheoreticallyMedia]

Yeah, I’ve got plenty of concerns on that front as well. I’ve long been a proponent of UBI, and that’s a conversation we should have been having 6 years ago, but per typical, we’ll be waiting until it’s too late to even start the discussion. That said, I don’t think it’s all doom and gloom. Although larger corporations will certainly downsize, this technology also offers many smaller companies to do more. All of whom will still require human talent to create. I think we’re moving toward less of a 9 to 5 cubical work structure and more into a remote/smaller team/smaller company situation. That said. That is on the knowledge worker front. We still have a much larger problem with automation (re: robots) in factory or labor jobs. Although not an AI issue, when was the last time you had a cashier ring you up, as opposed to self checkout, for example.

[@RSProduxx]

I can see it coming as in the games industry. A strong shift towards indy projects and small sized studios and I am not really against it. There is still bad quality/ good quality that will sort things out.

[@CelestialxPanda]

2:20 the only psycho is you here bro 😂😂😂😂

[@TheoreticallyMedia]

Haha, I see you’re one of those “light mode” people! You probably wake up in the morning with the sun shining in and greet the day. Meanwhile, we dark mode folks look like Gollum every morning!

[@TomasSowellIsGreat]

'Pun not intended' XD

[@TheoreticallyMedia]

Haha. It really wasn’t!

[@ReidKimball]

LOL! Why are they (MJ & SD) making this so public on Xwitter? All of that comms should be in private emails. smh

[@TheoreticallyMedia]

Agreed. I think it’s really amazing that David does the office hours every week- I can’t think of any other company that does that, and maybe this was just a bad week for him? But I agree, I don’t think this should have been aired before reaching out.

[@RSProduxx]

Man, I can´t wait for the day when I´ll be able to take all the stuff in my head and let it all out in short and longer movies :O Problem is, right now there are so many interesting tools for creative work, I´d need to be a good bit wealthier than I am to utilize a fraction of them properly xD Another thing that bothers me is that everything is online. I see censorship from the get go being a problem. Not necessarily talking about straight forward p*rn here, but other things including explicit visuals. For example, you want to create an episodic thriller series with some indecent s3xual or er*tic content. Where will those tools draw the line? possible outcome is that there´ll be dedicated tools for "adult content" which then again might be bad for "main stream" content or simply lack in quality. If you know what I mean... Well, maybe thinking too much about things we simply don´t know yet :)

[@fatmakhadija-gu3ic]

It only generated 2 or 4s video..for me

[@AG_before]

Thank you. 👍 Yeah, AI is stealing images from AI which got its... It just puts a smile on my face. :) Thanks again Tim. 👍

[@TheoreticallyMedia]

1000%! Always glad to see you in the comments, and really happy I can crack a smile on your day! Ok, off to research the next video! (Grind, grind, grind!)

[@_unrealgod_]

he looks like mark rober but older

[@pigeon-fd5zq]

Truth is that ai companies are now busy in hacking the code and dataset of sora ai

[@TheoreticallyMedia]

In one of the previous videos, I went over the paper that was reverse engineering Sora. So yeah, they’ve already cracked that. I think we’ll see something that looks like Sora before OAI actually releases Sora. Someone is going to beat them to the punch.

[@THEAMAZINGGREATCATSBY]

This app is useless to me I write stories and it's 3000 characters long but it only does 2 to 4 seconds I need something that does stories

[@TheoreticallyMedia]

Check this one out, for sure more in line with what you’re looking for: The Ultimate AI Film Platform: LTX Studio Deep Dive! kzbin.info/www/bejne/nKqakpWnoJdsaMk

[@firstcryptoetfnews4629]

U right

[@DeAnnaTroupe]

LOL at only psychopaths use light mode.

[@TheoreticallyMedia]

It's true!! haha

[@djnoj3371]

Thanks, I've been, using Haiper already but didn't pay attention to the SD mode. Just a heads up, my first two generations for a music video one Trump and 1 Biden, Trump came out perfectly, but it made a Biden a Black man.

[@TheoreticallyMedia]

Interesting...Yeah, looking around, I don't think Haiper cares on the IP/Famous Faces side. I saw a few Deadpools, a Mickey Mouse or two, and the Power Rangers. So...this is basically the worst fan film ever.

[@darakuT]

Are too short these video, 😢😢😢 only 4 sec

[@TheoreticallyMedia]

Yeah, that’s kind of the wall right now. It’ll be double that in a few months. Early days

[@Konz23]

Doesn't Haiper just do 2 Seconds movies 🙄 "Feature Length"

[@TheoreticallyMedia]

2 seconds in the HD mode, 4 seconds in the “standard” mode. The feature length refers to the second story in the video, mentioned in the intro.

[@DataJuggler]

I know it takes more resources, but until video is 15 seconds or longer, it's not that useful. 2 seconds reminds me of TV. I haven't owned a TV in about 15 years, and now if I see one, the screen changes every 2 seconds because that is attention span of the average person.

[@TheoreticallyMedia]

Yeah, I think the average shot length of most TV and Film is around 5 seconds. Which, is like, insanely fast. Personally, I think 8 seconds should suffice. For those longer "moody" shots, you could probably generate up 8 seconds and then slo-mo/extend them out. Did you ever see the breakdown on the Jaws One-er? (kzbin.info/www/bejne/bqKXiWWsebeZiM0)-- yeah, that is something that will be reserved for physical filmmaking.