Ranking The Top 13 Ai Video Generator Tools (Sora vs Pika vs Runway)

I've listed out Pika, Sora, Hotshot, Vo.

Just incredible to think that none of this is real. Like, I showed some of the videos to my son yesterday, and he was like, that's not real.

My benchmark up until now is I like to generate various monkeys on roller skates. I don't know why, but it passes my monkey on roller skate test pretty well. This one is definitely real. Nobody will ever know.

Hey, welcome to the Next Wave podcast. I'm Matt Wolfe. I'm here with Nathan Lanz, and today we're going to go down the AI video rabbit hole. We've recently seen Sora, we've seen Veo, We've had access to Pika and Luma and all sorts of AI video tools. Half of them Nathan and I have never even touched. So we decided let's bring on a guest who has played with all of them, put them all through the motions, knows the various differences between them. And today we've got Tim from one of my favorite YouTube channels, theoretically, media on the show, and. And we're gonna go down this AI video rabbit hole. So let's start with Pika. In my mind, Pika, where I think it really shines is, like, some of the new effects that they've put into it, like the ability to, like, cut, like, a cake or to, like, squish it or, like, the hydraulic press. Those are kind of, like, fun for, like, memes. And then they also. It is Pika, right, that just released that new feature where you can, like, put your own face into videos and then put yourself in, like, a certain sweater. I haven't really played with those features yet, but.

What do you feel is, like, Pika's winning features? Like, what would you go to Pika for?

I think that lately they do seem to be kind of moving into. Yeah, that squish it, cut it, cake it, explode it. It's kind of more on. It feels a lot more like it's on. They're aiming for, like, social kind of, like, fun tools, but I know that's, you know, you can upload recipes essentially, of, like, a person, a personal place, and a thing, and then, you know, stamp them all together, and then that'll create, you know, a video. But my understanding is that it's sort of on the template side. So once again, this is sort of falling into the area of, like, you know, just cool, quick videos to share on social.

Yeah. So with. With Pika, I really think that they've got, like, this sort of meme thing down. But when I was trying to play with the newest features in there, the ones where you can sort of add your own face and blend it with other styles and stuff like that. I was actually messing with it on a live stream earlier in the week, and I could not get it to generate what I wanted to. I was using, like, two images, and then somebody's like, oh, it doesn't work unless you actually give it a text prompt as well. So I was starting to blend it with text prompts, and even then, it was still giving me, like, really funky results, and I couldn't actually get it to generate what I wanted. I feel like Pika is one of those tools where I have to reroll so many times before I finally get what I want, but eventually I get there.

Yeah.

Yeah. I think they were the first one I saw that did, like, cartoon faces really well. Like, even, like, the expressions and the eyes. I think they were the first ones that kind of, you know, didn't perfectly nail it, but, like, at that point, it was way better than Runway. But since they came out, like all the other models that came out, and just visually, they're all more impressive, I think, like, most of the top models now, so I don't know where they go. Like, it almost feels like Pika should be some kind of new.

I'm leaning between B and C just because I struggle to see a lot of the real practical use cases outside of memeing. I just don't know whether it should fit in B or C. I don't think it's D tier because I think it's like a. It's og. They were one of the earlier ones. Them and Runway were kind of like the two big OGs that were putting out these video tools, and I do find some fun usefulness in it in that, like, meme thing. I've also used it at, like, the end of the videos where I'll be like, all right, guys, see you in the next one.

Yeah.

And then, like, the. The thing crushes me at the end, you know, so, like, yeah, there's some fun stuff that you can do with it. I'm kind of leaning towards B tier on that one.

Look, every video platform is, like, one, you know, update away from just becoming the greatest platform in the world. So, you know, that's.

Yeah, yeah, yeah, yeah. I mean, the last time we did a tier list, we. We definitely said some stuff to, like, save our butts so these companies don't hate us, and we'll do it again on this one. All of these tools. Yeah. Could be these right now, but one update away from being A's. And I feel that about every single one that we're going to talk about.

I could see Pika being like a really viral app at some point. If they keep going down the. The route they are now, or you put your own self into stuff, maybe it's not gonna be the thing that, like, Hollywood uses, but maybe it becomes like a really popular app with teenagers or something at some point. So I think. I think B makes sense because, like, right now it is actually useful for memes and stuff like that and like you said, like little parts of videos, whereas a lot of these, you know, tools, even though they look cool, they're not super useful in a practical way right now. So I'd say B.

Well, I'm going to move along here. The next one I want to talk about is Hotshot, because Hotshot, in my opinion, is like the D tier of the list for me. I don't know what you guys think if you play with Hotshot yet, but.

I could we see Hotshot, because I haven't seen it.

Yeah, Hotshot.

And for me, Hot Shot, I could not get it to generate anything that looked good for me. And they charged $99 a month for it, which is really high for what it is. And literally none of the generations I got it to do looked any good at all. But saying that we can move this around based on what Tim says, because Tim's probably played with it a little bit more than I have.

I have.

I have. I do. You know, those guys are. These guys are bootstrapping. They are. They're trying to, as best they can to. To get that model up and running. I do agree with you, though. A lot of the generations out of it always kind of have, like, this, to me, at least have this graininess to them, or there's a lot of morphine inconsistencies with them. And they were one of the first that kind of did the face swapping thing as well. But it always felt like my face kind of came out a little too much on the red side. Like, you know, the. The skin tones and color consistency weren't great.

So this is.

This is my Hotshot account here. And you can see I haven't done a ton of. But like, this was my monkey on roller skates here when I did a skateboard. Not even roller skates. This one here. Let's see blooming flowers emerging from melting snow. That's not too bad. But they don't really look like they're blooming.

Yeah.

Let me see here. Here we've got a ballerina pirouetting in a field of floating bubbles. I've got my speed controller thing turned on. So it's putting all these ones all over the screen. But obviously that doesn't look great. This is a chef juggling vegetables in a busy kitchen and the vegetables are just sort of floating around his head. And then astronaut planting a flag on a purple alien planet. The flag is just sort of like floating in front of him. And so I don't know, I was really struggling to get much out of Hotshot personally.

So this was, yeah, this was generated in Hotshot early on into their, you know, sort of face swapping thing. So that is a very unflapp, flattering shot of quote unquote me that did not quite make it. There was another one here. You know, this is obviously I'm a little bit better in the face there, but you can see clearly. You know, we get a lot of like backwards walking, a lot of morphing and inconsistencies in the detail in the background. And yeah, there's sort of another one where I'm kind of doing like an old man hobble. Like I, I'm walking like Harrison Ford actually walks now. And that's not very, not very heroic, is it? Yeah, I kind of like have an 80 year old man stumble there. So. So yeah, that's, you know, that's, that is my experience with Hotshot.

Look, prompting is all about technique. Just like any craft, using the right approach makes all the difference. The same goes for your AI prompts. Leave out the key details and you won't get the results you're after. Well, HubSpot's got your back. They worked with YouTuber and former metadata scientist Tina Huang to develop the ultimate guide to crafting the perfect AI prompt to using a three part framework. This framework will make sure that each prompt is chiseled to perfection so you can become a prompt engineer pro. Grab the guide at the link in the description below. Now back to the show. What model should we talk about next? I actually have not played a ton with cling, but I know everybody swears by how good cling is.

It can be.

I've generated maybe a couple videos, but I, I haven't played with it a ton, so I'm gonna have to lean on you a little bit for how good Kling is.

You know, I think between right now kind of leading at least on the Chinese models right now, I mean it's like neck and neck a While back I did a short film that was entirely AI generated called Dead Sea. It was a pirates versus vampires thing. It was kind of stupid, but it was fun. And that was all generated in CLING 1.0. So it began with a. So that's a mid journey image that is now animated in Cling. And if you go back and watch the. The original film, it's. It's clearly not as good as this. But yeah, 1.6 is doing some really cool stuff. Like the hands there look really great as he. I think the prompt here was like man telling stories, man telling ghost story or something like that. So it was a really basic prompt. Yeah, it looks, it looks really good and I don't know how much the resolution is coming through but like you can really. In his eyes you can, you can as. Even though they're sort of draped in shadow. Like you could actually see on a good enough monitor that he's kind of looking up and yeah, this, it's just interesting. I always like these sort of abstract experiments. This was another one that was like creative. I was like, just creative what it's going to do. It's kind of weird. Morphy and the model is much better than this. But I thought it was. This was interesting just to say what does creativity look like you to you model?

Yeah, yeah. I love doing like that super abstract prompts where it's just.

Yeah, yeah.

Let's see what the AI comes up with.

I think this was with their. Was one of the face training models as well. This, this was so, you know, compared to say Hotshot earlier, this is now much closer to me in my opinion. I mean, not. Yeah, I mean I'm nowhere near as heroic and cool as this guy is, but still like, you know, don't show.

This to your wife.

Yeah, exactly.

Yeah.

But she just laugh at it. Like, who do you think you are? You think you're that cool?

Like.

No, I do not.

So would you put CLING in like a tier? S tier? I mean I think we're talking top of the top of the chart here. So you know, if we're, if we're talking about what I would put in S tier, personally, I think, I think Veo Vo. Veo Vo. Is it like tier for me is, is where. Is where Veo's at. Maybe another one can make it. But that's where I would put it. Because the one thing that I do like about VAIO is that every time you give it a prompt it's going to generate four videos.

Right.

So you hear People going, oh, are these cherry picked? Well, kind of, but they built the platform sort of designed for cherry picking. They give you four and say, which one's your favorite of these four. Right? So, like, out of the four, almost every single time, I'm impressed by at least one of the four that it gives.

That is true. You know, I think. Yeah, I think it's. I think it is really good. I mean, again, the model itself has only been really released for a couple of days, so I think there's a lot of. That's. That's. That's a generation that looks really good, man. Yeah, I think there's some. That's really good. Yeah, there is definitely some. Some more work ahead of it, you know, like, it's not. So these are. These are your. Oh, that's great. These are your good generations, Matt Dog.

Yeah, these are. These are some of the better ones. I am kind of cherry picking some of my favorites right now, but I've got a lot of good generations. I also do have a lot of funky ones, too. Like, I tried to get it to generate like somebody doing a Rubik's Cube and definitely got some funkiness to it. Yeah, yeah. Not horrible. Definitely not horrible.

I mean, again, the fingers are not turning to spiders.

That is a win.

Yeah. It really sucks at gymnastics. Nothing has passed the gymnastics test yet that I found.

That's the new test after spaghetti.

Yeah.

The gymnastics is the new benchmark. This is probably my favorite generation is this rhino walker.

That's cool.

And it's like the camera's panning with it.

You know, My benchmark up until now was I like to generate various, like, monkeys on roller skates. I don't know why, but it passes my monkey on roller skate test pretty well. There's one of the monkeys on roller skates that I got. Here's another monkey on roller skates. This is the one that I shared on X that I thought was really good. I got it to do like a first person looking out of a jet fighter.

Here's another one.

And another one.

It's just incredible to think that none of this is real. Like, I showed some of the videos to my son yesterday from vo and he was like, that's not real. He's like, oh, that one's not real. That one's not real. And then there was one with, like, an animal or something. He's like, that's not real. No, it's not real.

Disappointed.

Yeah, this one is definitely real. This is a total.

Nobody will ever know.

NASA just announced this today that they just.

Yeah, they just put flying.

They forget what the drones were. That was. That was. That was the drones right there.

Let's see. One of the ones that really did not work out well was. I tried to tell it to make a guy playing with his dog with dolphins jumping in the background, and that's like, barely even a video. Here's another one that it did. And the dolphins are just floating there in the background. So it's not always, like, the best thing you've ever seen. Right. Like, there's definitely some issues with some of them. But again, every single time it gives you four, and, like, one of the four is usually pretty good.

Were you doing. Was that a lot of the image to video or is that more text to video?

That was all text to video.

Text to video. Yeah. I do find that it tends to generate a lot less wonk when you're doing text to video as opposed to image.

And one thing about veo too, is when you do image to video, you can't upload your own image. You have to generate an image with their image generator.

Yes.

And then you can turn the image that they generated for you into a video. So I think it's using their image in three. Right.

Which looks good. I got to say. That's actually. I can't really complain about, like, that. Imagen 3. Yeah. Has taken a pretty big jump up from Imagen 2. It actually looks pretty solid, in all honesty.

Yeah, it is good. And it's got people really well. Yeah.

I haven't been able to try this yet. So, like, in terms of a product, like, how do you. How does it compare to, like, you know, Runway? It's got a pretty great editor tool. Sora now has a pretty awesome storyboard.

It's definitely still a beta product. It's definitely still a beta product. You know, like Tim mentioned, it does not seem to have, like, any sort of library. So anything you generate, if you refresh your page and you didn't download those videos, those generations are gone. You can't find them again. It doesn't really have any editing features yet. It doesn't have the ability to upload your own images and convert them to videos. It's definitely, definitely beta. Yeah.

That's.

Yeah.

That's why I have mixed feelings on, like, on the S. Because, you know, Google has done this so many times where, like, they have amazing technologies.

I know.

You know, you called me cynical the other day, but, like, I kind of am about Google. I know. I know I am. But they haven't really launched great products though. So even though they have great technology, that doesn't mean it will turn into a consumer product that people actually use.

Yeah, no, I agree. Like, I think that's where the debate would live for Veo. Right. It's like as far as the video generation technology goes, I think it's the best right now. I think that's the state of the art right now. As far as the fact that most people don't have access to it yet, who knows when more people will get access to it. Who knows what they're going to charge for it. Who knows if they're going to build a good UI around it. There's still a lot of questions around it which, you know, I could see the arguments for knocking it down a tier because of those reasons.

Yeah, yeah. I mean Sora has like a, you know, great domain store.com sexy landing page, looks beautiful. It's got the storyboard feature. I mean those are all, you know, those matter for like regular people.

So. Yeah.

But some of the generations are definitely looking really, really good out of it as far as their overall plans. I mean I think that we will definitely, I think we're going to get a definite release out of it. The. There's already been talk about YouTube integration with it and then Matt, I don't know if you saw as well, but did a little banner on your channel pop up that asked if. If you will allow for your videos.

To be used as it might have. I might have closed it real quick and didn't pay any attention to it. But I, I knew that. I know they have that feature just rolled out where you can decide whether or not you want to allow your channel to be trained on or not.

Which I figure, I mean, I'll turn around is fair game. I mean I will turn it on. I will. I mean like look, I've like who wants to prompt a guy, you know, in a studio talking to a camera with a bunch of like mid journey images and AI video? You're going to train off of AI video, which, you know, whatever, you know.

But I just, I feel like they had to have already trained on YouTube video.

Maybe I'm wrong.

They definitely did.

This is like now like covering their asses after the fact. Like hey, do you opt into this.

No matter when we did it?

I have been desperately trying. That's another one of the tests that I, another one of the tests that I really love to do is to, is to try to prompt for me. So I, you know, I've Run that into chat GPT number times, just a screenshot of me and say like describe everything that you see here and make me a video or image prompt out of it. And I'm constantly running that just to see like how close it gets. So far I haven't gotten me. But I'm also not Marquis Brownlee, so I'm not that level of famous.

So yeah, I do think they will roll it out. I bet you it'll. It'll roll out like into Gemini or something where you just sort of can generate videos in line with your chat and things like that.

Kind of like Dolly. Dolly and chatgpt.

Yeah, yeah, I think, I think that's probably where they'll go with it. But I do want to keep on moving along here because we're about a third of the way through these tools here. So.

Okay, they still get an S. They still get an S because they like video wise it's the best quality.

So yeah, I'll leave them in S just for that reason. But moving on, let's talk about Sora because Sora and.

Sort of. Yeah, I mean, but let's just go ahead and get that one over with because I know we probably have a lot to say about that one. The thing about Sora, and here's where I'm struggling to place Sora, what we have access to right now is Sora Turbo, right? And Sora Turbo gets us 22nd generations. They've got some additional frame interpolation built into it so that you don't have to generate as many frames. But it's also makes it a little less smooth with how the videos come out. But then you also have like Sora Regular, which is like some of the demos we saw nine months ago or whatever. And those look a hell of a lot better than what we're getting out of Sora Turbo. So it's like how do we rank it? Are we ranking it based on what we have access to now? I think you gotta rank what we've seen from Sora.

I think that you have to rank it based on what was released and what is available. And to that, I mean like I listen, I know that it def. There's a, There's a large segment of the, of the, of the population of the AI video generation or video population that kind of. It landed with a wet fart, you know, like people were not stoked about it. But I also think that there is still something there. Is it worth $200 a month for most people? Probably not. It probably is one of those things where everybody came in with their standard sort of prompt structure, tried it out, it didn't work, or it was just like weird, wonky results. I think that we just have to spend more time digging into it and figuring out what it's really good at. There. I did the video to video aspect. The remix and the blend and the fact that you can cut, there's a lot of power in that. Probably those are features that we haven't seen in any other video model. So yeah, there is really good stuff in there. Again, I know the controversy side of it is just like you can't generate people on the, on the $20 plan. Yeah, there's a lot of. You can't, you can't, you can't unless you pay $200 to.

I have not gotten anything that I'm super impressed with out of Sora, if I'm being honest. You know, here's my monkey on roller skates. That was probably the best prompt that I got out of Sora. But then my VO Veo prompt with a monkey on roller skates, like blew this one out of the water.

Yeah, yeah.

I asked it to do a wolf howling at the moon and. And it literally just gave me like images of wolves howling at the moon. Not even a video. And then I started to learn. All right, you need to give it more detail in your prompts if you really want to get something exciting out of it. So when I started, you know, putting a lot more detail into these prompts, like you can see how long of a prompt that is on my screen. The video started coming out quite a bit better. But still, I don't think they're on par with, with what we're getting out of Veo, if I'm being honest.

Yeah, no, I would agree with that. And that I think definitely comes down to training data. I mean, obviously Google likely has access to the entirety of YouTube and you know, so. Yes or. Well, if you believe Mira Morati did not have access to all of YouTube. They did, but yeah, or, you know, Turbo might also. That's actually kind of cool. Turbo actually might not have as much of that data in there.

I think the product itself though is probably the best looking AI video product though, in terms of like the user interface 100% and how they present it.

Yep, there is definitely a very mid Journey esque aspect to the to when you first log in and see that sort of endless scroll. But yeah, in that, you know, again, the remix, just all of like the little pull downs, the simplicity of the Prompt like, no, it looks great. I think again, you know, this is one of those things where I always try to be a little gracious with them when they release a V1 product. And, you know, essentially we're beta testing their V1 right now. By the time this hits. Like, again, remember how weird and wonky Midjourney V1 looked to where it looks now? So what will Sora 6.1 or 7 look like? Spider 1 is cool.

Yeah. I pretty much did all of the exact same prompts with Sora that they did with like Veil, because I wanted to compare it with those.

Yeah.

And for the most part, like Veyo won on every single generation.

Yeah.

So I can't put what we currently have from Sora in S tier. Maybe what I've seen from Sora, some of the videos they gave us six months ago, nine months ago, some of those might fit in S tier. But what we've got now, I think what it's got going for it is the UI is really cool. The storyboard feature is really, really cool. I like that a lot. But at the end of the day, it's like most of my generations, I don't feel like I would actually use as like a B roll over one of my other videos or something like that.

I still feel like it's like looks like the second best. I don't know, maybe like the top of A or something like that.

Well, I haven't used CLING a whole lot, so it's like, how would you like if you were ranking veo, Cling and Sora, like, how would you rank those?

If I was to do all three right now for my use cases and for sort of like the pack that I travel with, it would actually probably be CLING on the S tier. Between those three would probably be CLING on the S tier only because you could generate images from outside of.

The platform. So on Veyo, you have to generate in or veo, you have to generate in Imagen, whereas in cling, they're like, you can generate in stable diffusion, you can generate in flux, you can generate in midjourney. They don't care. It's just that you bring it in and from there you have motion brushing and all of your other various tools. So there's a lot more control of what you're choosing to input in with Cling than there is with, you know, VEO right now or Sora, you can actually, you can actually bring your own images into as well. I'm noticing it, it's still going to take A little bit more testing. But they tend to begin with your initial image first, and then they kind of wander off into their own world. Like, the first frame will be like, you know, oh, this is exactly what you're looking for. But I'm going to do something complete, not completely different, like something that's based off of what I see here. But it's not going to be. That's not going to be the first frame. I'm not going to continue on with that. Which, I mean, that may be just the new way of. Of. Of generating a video. Letting the.

I do have an example of that.

Yeah.

Let me show you this example. So here's one where I. This was like a YouTube thumbnail sort of starter image, right? And I gave that the. The video and then I just gave it the prompt. A man speaks into a microphone. And if you watch this video here, as it eventually plays, you'll see that it goes, cool, nice image. But I'm just gonna use a different image completely. Yeah, the one on the right.

Copy skid or something.

That was a nice image. But here's what. I'm gonna actually make a video from it, just which is the image out on me.

But it's still. But it still stays you, though. Like, it still stays, or like the character of you, it just goes. Moves into a photorealistic direction. And I think that might be something that we're looking at. And I know that that's going to be controversial to some, but in some ways that the video. Those, like this, this wave of video models is kind of saying, like, listen, we're the director and we're the production team here. Like, you're writing this, but we're in charge of actually making the thing. Um, so it's almost like you, you're sort of like a producer or a writer and, you know, you're handing out a script to a production company in, like, Bulgaria and yet you don't know what's going on. They're just sending you back stuff and you're like, well, this isn't exactly what I asked for, but I guess it kind of works. So we'll see. We'll see. You know, and I would say. The other thing that I would definitely say on the SORA side is that was a. Was a prompting part of it. It is weird because I do think again, like, Matt, like you said, you need those longer prompts. Whereas with, like, as we have seen with. Well, we'll look at Minimax later on and with Kling, there are sort of those AI enhancement. You know, like tools that are built in there that you can do something to a monkey riding a skateboards. And then the. The model will fill in a lot of the additional details that it needs to generate the image. It is so weird that OpenAI did not put that in. Like, why isn't ChatGPT right in there? So I don't know.

Yeah. So, yeah, Sora Turbo feels kind of like, rushed out. Like, I think they were pressured because, like, the public, you know, consensus was like, what the hell is going on with Sora? It looks so cool. And we haven't seen anything. They don't ship anything.

Yeah.

So then they had to get like a cost efficienc IT model out as quick as possible. And it's.

And it did leak like a week earlier and got. And a bunch of people managed to get access to it a little bit early, and then they had their 12 days of Christmas and it was like, all right, well, I guess this is probably the opportunity to push it out, but yeah, for sure. Well, you mentioned Haylou AI let's like. Or is it Min Max? Minimax?

I always just call it Minimax. Like, it's just because that's Hailu, I think, is the. Again, the actual name of the. And then Mini Max, I think is the name of the. Of the. Of the video model. So.

Yeah, got you.

Yeah.

Yeah. So I always. I always call it Halloween, but I always just do it almost to, like, make fun of it when I make videos. Right. Like, I just sort of screw up the name on purpose because I think it's funny. But so where. Where would you rank Min Max or Mini Max? Like, I've. That's one I have not played with yet at all.

Oh, really? Literally, the. My entire audience would come after me with pitchforks if I did not say that was S tier. Uh, Minimax is definitely the one that I think is, like, current crowd favorite. Everybody loves it. I love it too. It's. It's a great model.

Would you put it above vaio?

Yeah, currently. Yeah, for sure.

Really? So, see, I haven't. I haven't played with Minimax, but, like, all the examples I've seen, like, all the demos, like, I don't know, like, VAIO seems like way better than Minimax or maybe.

Well, again, with Minimax, I think that the.

The trick, you can use external images and stuff, I assume.

Yeah, it's the control. Like, you know, there's like, there's so much control in. And prompt adherence.

Yeah. I mean, they're not as Worried about being sued? Like, definitely. That's a problem that Google and OpenAI have, is that, you know, making sure they're not getting sued. You can't just, like, copy people's faces and make entire films out of it.

So, yeah, these were. I was running. Yeah. So, you know, bad tests here, but I was running the Sora Viking thing, trying to get that kind of look out of it. And again, I mean, it's like, it's, you know, compared to that Sora leak demo that we saw earlier. Yeah, of course it's not going to be as great, but like, that looks. That looks pretty good, you know, for image to video. Yeah, pretty solid. You know, it still does the walk cycle thing that you can kind of see that in her legs, but still, it looks like that looks pretty smooth and good. The other thing that I think it's really good at, I just did a video not too long ago. It was about an interview. So it was this guy and. And this guy. So it was a job interview. So this was our interviewer. And then. Yeah, this was a hitman that he was hiring or that was going out for a job interview. And, you know, the running joke there is like, what kind of experience do you have? And he's like, well, I'm very patient, you know.

So. And I used Runways Act 1 to do all of, and then 11 labs to do all the voices. So this kind of stuff, it's really, really good at. And then we had like, various flashbacks. He's like, I have international experience. My. My wife was killed in Thailand. We meant to run away from this life together, but. And then the triads killed her, you know, like, so. So this is the kind of stuff that I think it does extremely well. Like there's another flashback of like, his CEO, when he was in Special Forces, was a traitor, and so he had to have him executed, you know. Yeah, this, you know, for the kind of doing these short clips, like, there's like. I mean, like, yeah, Minimax is. Is really on point. It really does a great job of understanding exactly what you're looking for. A lot of these are running, by the way. No prompt. Like, I'm just like. I'm not even putting in a prompt. It just. It's like, oh, you have a city there. Yeah, I know you kind of want to do a thing. No talking, minimal head movement. You know, you don't have to go crazy with your prompts. Like, I think the girlfriend here was no prompt. It knows waves, it knows, you know, move her around. Yeah. So this is the kind of stuff that like, no prompt.

Like, so are you generating images and then pulling them in? Is that what's going on?

Yeah, so this is, this is all generated. The, the initial images are generated up in mid journey. That's usually how I tend to midjourney or flux. I'll go back and forth on, on either of those and then drop them in. And for the most part, I mean, yeah, Hailey will always understand what you're kind of looking for. Yeah, you're going to get a couple of like weird morphe things back here. Like these guys kind of do a zombie shuffle as they're kind of running towards him. But you know, if you, if you run this a few more times, you can usually get that to work. This was man on trial. Was this the bad one? There was one where it just kind of. Oh, no, that was actually the good one. There was another one. Oh, this one. So this is a good example of like, sometimes it just does not do what you wanted to was like he just moves off this way and the camera just kind of goes off in a different direction. So that happens from time to time.

But yeah, I mean, I'm, I'm looking at their website right now and all the ones that are on the Explore page are really impressive. Like, yes, like, better than mayo, honestly. Like really aesthetically pleasing. The lighting looks great, the colors are great. Like, I don't know, I have to actually kind of agree just looking at the Explore page, that it's really impressive.

Yeah, I'm kind of surprised by this. I had seen it before, but like, this is better than I remember.

And like, again, a lot of the prompts here that you can see in these demos are very. And again, you can kind of go either way always, you know, prompt if you're not getting the results that you want to. But a lot of this stuff is very. It almost kind of looks like this prompt here is actually just the image prompt. You know what I mean? Like, and then this, like that was the prompt they ran in mid journey and then they dropped it into here and just kind of reran the prompt in here to kind of reinforce the idea.

No, it's S tier or.

Yeah, let me lean in. S tier. I'm leaning. S tier.

Yeah, top of S tier or.

Yeah, right now I've got that top of S tier, but it's right next to Vea. Like they're, they're, they're, they're sharing that, that tier together. All right, so let's, let's talk About Nvidia. Have you played around with Nvidia much? And Video is one of my sponsors for my other channels.

So it's hard for me to like.

I'll put in a D tier.

I think they've offered to sponsor me a couple of times as well. I really haven't got. Not gotten the chance to like, I dove into them I think early when they first launched and I really haven't gotten a chance to circle back to them. So yeah, that's, that's one that I'm not. Yeah, I don't really play around with that much.

Yeah. So within Video, one thing that like makes in video stand out is you give it a concept for like a short film, right. And you can actually say generate a five minute video for me and it will actually go and generate all of the scenes for your video similar to what you can do in like LTX Studio. But yeah, you can. You basically say like I want to make a movie about a Viking, you know, hitting the shores and then attacking the. The locals or something, right? Yeah. And it will go and generate like a three minute video of that happening and each scene in it is generated, I don't know 100% if InVideo is using an API or if it is their own internal model. I believe it is their own internal model. It's pretty good. What I actually use Invideo for more and like it for even better is it's got like stock footage and so you can give it a concept for a video and it will go find stock footage for that video. I actually think it's stock footage features are really where Nvidia shines more than its generative features at the moment. But saying that, you know, they're only like one update away from, you know, jumping up the tier list. But right now, just speaking about like the generative capabilities of it, I'd have to probably put it in B or C. But the stock footage feature is really, really, really good. It's really good at going and finding stock footage for whatever you give it.

And it's funny because I remember the video that I did, they got me for a sponsored spot a while back when they first launched. So this was before they were doing generative stuff and it was just the stock image stuff. And of course me being me, I wanted to push it into weird directions and see what we could do with it. So I ended up making a. It was a zombie survival guide, but it was picking everything from stock footage. And I generated up the script. It was like a British Guide, like how to survive the first 10 days of a zombie apocalypse and is, you know, it was very kind of like, I guess like YouTube explainer y. But it was. Yeah, it was a neat idea. So you can kind of like weirdly thread narratives out of stock footage as well. So it's all just a matter of kind of how you approach it. I thought it did a serviceable job, but again, I just haven't gone back and checked it out from.

From a.

From a generative standpoint since.

Yeah, and it also does, like, it'll write the story for you, it'll do the voiceover for you. You can upload your own voice and actually have it use your voice in the video. You can do all that kind of stuff. So it's got some really, really cool features, but the generative capabilities, I mean, I don't think I can put it. The generative capabilities on the same level as any of these other ones. So I think I would probably put it in like the lower B tier right now, possibly bordering on C tier. But not. Not to say that Nvidia is not a great product. It's just. I feel like the generative functionality has a ways to go.

Oh, I've got a. I've got a. I just thought about this. I got a controversial S tier coming up. Okay. Yeah. Okay, well, let's jump till the end.

Oh, should we leave that one to the end?

It's. Whatever you want to do, it's fine.

All right, so which one would you put into S tier next?

Oh, S tier. So this is going to be controversial, but especially coming off of the Nvidia one, I'm going to put LTX in S tier. Why? Okay, like, aside from this, they do that. We both, Matt and I are sponsored by them. But the. It's actually, it's not that they're not. No one's putting that money in my pocket right now. No, literally, because they open source their video model. That is. Like, is there any other company here that open source their own video model and gave it to the community?

Not of this list, I don't believe.

So Meta didn't do it.

Again, I don't believe any of those are open source other than ltx. No.

So, I mean, that is.

Do you think LTX from the generative side of it, though, do you. Do you think LTX generates videos on the same level as what you'd get out of Minimax or veo?

Not currently, but again, the code is out there and people can tweak it and people can do things with it to the open source community, can bring it up to speeds or to a quality of that level or perhaps even higher. They can do, they can tinker with it, you know, because again now you've got a worldwide, you know, network of people that can now futz with that model to make it do kind of whatever they want. And it actually does generate extremely fast. I mean, I think that's the selling point of the whole thing is that it actually generates sounds great in theory.

But like massive compute to like advance these models in the next year.

Yeah, I mean I think it'll take open source.

Sounds nice but like how are they going to compete with like Soar Microsoft.

The training data and all of that stuff and all of that. I mean, you know, again, I think it's, it's not necessarily for the, for the, for the massive ramifications that, but just in the fact that it's like, you know, it gave, they gave their model away to everyone. That's, that's actually again, that's just super cool and giving back to, you know, a community of developers that ended up bringing, you know, bringing all of this technology together.

So yeah, I think that's good but I think we should try to judge it based on like how useful it would be to people.

I am sort of leaning more towards Nathan's take on it. Like I see, I see why it goes up the chain because of the open source. But at the same time, if we're judging, you know, if.

OpenAI and Sora based on like what we're getting today right now, I feel like we should be judging LTX on, on what we get today, right now from it.

Yeah, if we were judging the future, I would literally put SORA at the top of the top, honestly would be my personal take.

Right, right, right, yeah, no, that's valid. I, Whatever you guys landed, tier wise, I think it should just bump up one for, for the fact that, you know, they gave.

I would have probably put the generative capabilities on par with what you'd get out of like Nvidia or Pika. But because it's open source, my I, I say it bumps up a tier and gets an A. I'll say.

Have you been around lately? Because they've actually been doing a lot of improvements in there as well. The video model does look a lot better, a lot smoother. I know it was a little bit on the rocky side to start. They added flux in as their image generator too, so that tends to look a lot better and I think Today they kind of put in. Was it sort of like that advanced live portrait thing where you could kind of do expression control and everything. It's funny talking to those guys that it's. They have such an interesting problem in that they are kind of a do it all platform, right? And as like new technology kind of appears, they're kind of like putting it all together. But it's not just a matter of just plugging it in. It's not like, oh, here's a module that we just plug right in. It's almost like, you know, you've got this existing. It's not like a car where you're like, oh, I'm going to drop a new. Or actually, you know, a good example is like, it's not a computer where you're just dropping in like a new RAM slot into it. It is more like an 18 ton semi where you're like trying to drop a new engine into it or you know what I mean? Like, like a new like catalytic converter or something. So there's a lot of like playing around to get the whole thing to work together. But I think they do a pretty solid job of trying to stay up on everything and kind of creating that all in one platform.

So then they also have LTX Studio, which does similar things to what we talked about with Nvidia, where you can give it an idea for a video, right? And then with that idea, it will actually generate all of the scenes for you and then you can go in and swap out scenes. It'll do the voiceovers. It'll essentially make like a little short film for you based on just like a prompt, which I think, you know, that feature sort of bumps it up a little bit as well.

I think where people get sassy with that too is that, you know, they, they kind of go in there like maybe a serial killer werewolf meets a alien, you know, vampire robot story.

My exact.

And then, you know. Yeah, which I would, I would totally watch. That sounds like something I would. I definitely watch when I was 13 on Cinemax. And then they get something that is like it's kind of there and it's kind of not like it just in the video, the, the image says that they pick aren't great. Like it is that thing where you've got to spend a lot of time in there. It's not just like prompt and like I'm instantly entertained. It's like it's prompt and then you got to start going in and you know, changing things on a granular level. I think that that's where people are just like, oh, it's not that good. And it's like, no, it is. You just have to spend a lot of time working on it. It's not going to, you know, you're still going to have to do the work, basically.

So Luma. So we've got Luma Dream Machine. Where would you rank Luma Dream Machine?

I really like the generations that Luma Dream Machine comes up with, but I don't think it's on par with what we've seen from Vaio or from Minimax. I'm curious where you would rank it based on all of your testing.

But the new image model looks really good. So you can now generate on platform. They're kind of doing something a little bit on the similar side to. It's not quite ltx, but kind of more like Leonardo's flow state, I guess, where you can kind of have a bit of a conversation. It doesn't quite do like that cascading like level of like of images that keep going down. But you, you kind of end up. You Remember when Dall E3 first came out and you were like, you were supposedly be able to. To chat with your image and, and you know, and get it to do things. That's kind of the. That idea in. Yeah, it looks really good. You got to give it up because their APIs everywhere, anywhere that you go, you can now generate in Luma, the first frame, last frame thing you can do some really cool things with. I think that they are. I would probably put them into the A category somewhere. Has somewhere between A and B. Like they're. They're just. But they're just like one update away from like hopping up to an.

Most of what I do with them is the. The image to video. I. I feel like with this one, starting with an image, you tend to get a much better result than if you start. If you just give it a video prompt. Like here's one where I put a video or an image of me and a few other people at an event and then told it to make us dance. One of those guys, that's Don Allen, that's Roberto Nixon. This guy runs chatgpt tricks on Instagram. But, you know, they all. They're all pretty good. I don't know what's going on there. Sorry about the neck beard.

But you know, the other thing. Let's see. This is. Am I thinking of the right tool? This is the one where you can give it a start and an end frame, right?

Yes. I've Got some cool.

So you can give it like a starting image and an ending image and it will sort of animate between the two images, right? Yeah.

I'll show you something that I was playing around with this idea about wanting to do kind of a Game of Thrones type opening. So, you know, I started off by generating up some, you know, like, still images of like these castle things that wasn't quite. It ended up kind of arguing with. With Luma a bit. Ended up zooming out, you know, with. Once we established a look, zoomed out a little further, kind of got into a map look. And then, you know, from there you could kind of start taking it into. Let me see. It was one of the 12 second ones.

What's this user interface? Is this Luma?

This is. Yeah, this is Luma now. So, you know, we can kind of get these kind of cool effects using first frame, last frame, you know what I mean? Like, kind of getting this real sort of like swooping, droney, not quite Game of Thrones, but, you know, kind of like, you know, by the end of it, I had like, you know, fairly quickly too. I ended up with like this 19 second shot of like this, you know, fly through. I think this might be the one where we go into the window and then into the throne room. Yeah. So there's kind of like that weird stutter there that I would have to fix in post. But now we're kind of flying into a throne room and now we got this guy and eventually actually in the final version, it actually ends up coming out, we go through him and then everything explodes into fire. So, yeah, this kind of stuff is like, really fun and very easy to do in. In Luma. So. Yeah. Yeah, I think Luma is a great little. It's a great exploration. Yeah, it's great for exploring. I actually played around with an anime style too, which I normally do not do on my channel.

Yeah, I'd say Luma for me is probably a tier. I really like that sort of transition feature where you give it two images and then it can transition between them. I think that's really, really handy. So for me, I say Luma's a tier.

Yeah, for sure.

And then like, so moving along here, next up, we got Runway. And Runway, in my mind is like the OG right when it comes to, like video generators. They were out there doing it first. Right. We had Gen one, Gen two, Gen one.

All those three seconds of video, my friend, three seconds of video. And people complained.

So it's like you can't put them In D tier, just because they're like the godfathers of, like, the AI video.

Tools, you know, maybe hot take on this. But I would actually, I actually even. They're one that I would put up in the S tier only because, again, this suite of other tools that they have as well. I mean, Gen 3 is great and it's fast. I mean, it is super fast. I don't know if you've played around that much with it on the image to video stuff, but you add Act 1 into there and like the expand and all of that other stuff that they've been doing lately. I mean, it is a killer suite of. And then on top of it, you can just take all of that stuff and instantly rotoscope and green screen in there. And like, they've got a subtitle thing. If you go into, like, the greater.

They keep adding new things, right? Like, I always see new features they're releasing.

And then if you go into the greater, like, Runway ecosystem, like, there's just so many tools down there as well.

Don't they have over 36 tools, right? You got audio, lip sync, moving backgrounds, text to images built in. I mean, one of the guys that runs Runway used to, like, he's one of the original creators of Stable Diffusion as well. Um, so it's like the. They're OGs that created Runway. They go way back to, you know, the very beginning of AI image generation. And yeah, there's just so many tools. Like, I use the green screen tool quite a bit where you could take another video and remove the background. I think that's really, really a handy tool. But yeah, when it. When it comes to like, act one, where you can just sort of upload a video of you talking and change it into, like, any sort of cartoon.

That'S super cool, you can now do it with. I really think it should have been called an Act 2 as well. Now you can drive that Act 1 video footage onto any other, like, footage as well. So you can have somebody walking around in a 16,9 frame and then just completely change, you know, essentially what they're saying in their facial expressions. Like, that's super handy. Like, it's huge.

Yeah, well, I mean, you got all these camera controls, right? Horizontal vertical zoom, roll, tilt pan. You know, you've got expand video where you can give it like a vertical video and turn it into a 16 by 9 and have it actually stretch it out for you or vice versa. You know, it's got a ton of cool features. It's really, really good at image to video as well. Like you mentioned, it's fast, too. Let me see if I can just find something real quick to drop in here and see how fast it generates. Oh, and this one has a first and last frame, too. I didn't even realize they've added that feature to Runway. So here's like a colorful swirly one. And if you miss pretty quick, if.

You miss a Friday, that seems to be when Runway loves to drop an update. So you always got to pay attention.

Yeah. Yeah. Well, I always make my news videos on Thursdays and release them on Fridays. So news is always like. Yeah, I'm always talking about it a week behind.

So it sounds like Runaway's got the best editing tools now. Right? Like the best editing tools. But I still feel like the visuals are not the best. I mean, they're good, but they're not the best of the best. Is that a fair take?

I. I think depending on what you're. What you're generating a lot of times, like that one, the, the matte one here looks pretty good. Like. And again, you do have all good. Your video controls, your. Your motion controls and all of that stuff. And depending on what. Because this was promptless too, right, Matt? So, you know, there's. I think it depends on what you're looking for. And then on top of it, because you can take this up to whatever, 10 seconds. And there's all kinds of other tricks. There's like, I'm always thinking a lot in terms of, like, you know, I, you know, use all of the meat of the animal. So when I see a generation like this, what I'm thinking is like, okay, I'm gonna bring it in, I'm gonna run it through Topaz, and then maybe I'll speed it up. And I'm only going to take the first, like four seconds of it to give it a quicker vibe or quicker feel to it. Maybe put a zoom in. And the, you know, I'm just thinking about all the other ways that I'm gonna use this. Not necessarily just the generation itself.

Yeah. And that's another thing that when it comes to, like, generating videos, a lot of people don't really think about that is in most scenarios, you really only need about three seconds. Right. Like, if you ever turn on your TV and watch a TV show or you watch a movie or something like that, they never linger on a shot for more than three to four seconds. Right. So if you can get these tools generating 8, 10, 20 second videos for you, if just 3 to 4 seconds of that video is good. You've got perfect B roll, right?

That's something that when that Sora thing, when SORA was first announced and was like the 1080p 1 minute long generations, it's like no one's ever gonna. You're never gonna watch one minute of like, you know what I mean? Like people can't. Can barely hold their attention through a one minute TikTok or YouTube short where it is cutting like crazy. Like try making a, a YouTube short one day when it's just you literally just talking to the screen for a minute and like, I guarantee you it'll be the lowest watch short of all time.

100 so runaway. I don't my problem with S tiers, I feel like visuals, it's like B. Like in terms of if I was just looking like how good do the videos look like? Like recently I put together like some Twitter threads like comparing and. And I didn't actually include many videos from Runway because I just didn't think they looked as good. I didn't think they would perform as well in the thread. And I, you know, when I looked at all the different models, I mean it was like, you know, VEO was by far the best. And then Soar was pretty good and I felt like Runway was like a, like slightly a tier below visual quality wise. But the editing tools are amazing.

And it's not the most controllability though. Like if you saw like the camera angles and the like. From a controllability standpoint, I think Runway is up there whether I put it on the same level as what I saw from Minimax and from vaio. So then, all right, we've got two last tools and these are the two that we don't have access to. I mean technically I think all three of us actually have access to Adobe Firefly video, but the emails I've gotten from Adobe says I'm not allowed to share those videos yet. So.

Okay, so D tier.

Yeah.

Really?

I thought the model was released and out to the public.

No, the Firefly video model, I literally got an email from them this morning saying that you can play with it, but make sure that you send any videos to us for review before showing them off. Yeah, so what I figured we can do for this is we've got the.

Current model that's out probably is what we should do.

So for Adobe Firefly, we've got their page here with some probably pretty cherry picks demos. But you know, you've got like this dog video here and it seems to be one of the better Ones maybe with text from, from what I've seen. But, you know, some of this isn't, isn't great. Some of it is pretty good. You know, based on what you've tested with Firefly, we can't actually show any videos that we generated because of the rules there. But based on like the videos that you've tested, like, how would you rank it?

I, you know, I think it is very, I think it's very Adobe. Like, I think that if you are, you know, working as, Yeah, I think if you're working as a professional editor in you are, you know, you need some B roll of a thing. You know, much in the same way, like any. How much are professional photographers using generative fill? Like, is my, you know what I mean? Like, I think that it's become a tool that they, they'll try out a few times. They figured out where the use cases work really best. They know where it, it doesn't do great. Like, you know, if you're a wedding photographer and you need, you know, I don't know, something like a couple of birds up in a tree or something like that, I'm sure that, you know, Firefly or genfil is going to do that for you. If you need dragons flying in and attacking the, the wedding party, it's probably not the tool that you, that you're going to use for it. So. Yeah, yeah, I feel that that Firefly video is kind of in the same, in the same area if you're doing, you know, kind of like what I would call like regular movies or regular videos, like, kind of like very, like, I don't want to say, I guess stock ish kind of stuff. Like, yeah, it'll, it'll perform very well if you're trying to get into doing your, you know, if you're trying to make Avatar for like, you know, then you're going to struggle with it.

So, so the other thing that I've noticed about Adobe and this is more with the Firefly image generation, but I imagine it translates to the video as well, is as far as like, censorship goes, they're the worst. Like, they're the ones that like I tried to generate an image of a person standing in front of the Eiffel Tower and it's like, oh, we can't generate the Eiffel Tower. Right. Interesting. There was like, lots of things that it would not generate for me that it didn't make sense that it wouldn't generate.

It is always hard for me to argue against. Like, because, I mean, Adobe is Obviously, very much taking a hard stance on, like, we are trying to be as ethical as possible. And it is sometimes hard for me to be like, well, stop.

So I'll do it. I'll say stop.

But. But yeah, we let Nathan be the.

Bad guy on this show.

Yes, I have no hards, cynical and just stop.

So, yeah, I think that that is. Eventually, I think that that will be. But again, I think that where Adobe, where Firefly and Gen Phil really do shine, is the fact that they are going to be directly integrated into Premiere into, you know, into Photoshop, as we've seen. And that is, you know, so it becomes that at least first where you're stuck on a thing, you know, at some point when this, like I edit in Premiere at some point, you know, when I need a quick shot, like, am I going to bounce out necessarily to another video model while I'm editing? Like, if, if I'm like, oh, this just occurred to me, I want to try this out, or am I going to give it a shot real quick with the new Firefly video model just because I'm already there, you know? Yeah, yeah, yeah.

The other thing that I really like that Firefly is going to do, I don't believe we have it yet, is that generative extend, right where you have a video that you need it to cover eight seconds, but right now it only covers six seconds. You can, you can extend it out by two more seconds and it figures out what the rest of that video would have looked like. Right now, when we're doing video editing, I don't know if you're like me, but what I'll do is I'll literally slow down the clip so that it's a slightly more slow motion to fill that gap. But I like that idea of having that generative fill inside of video to sort of fill out that extra one or two seconds that I need.

Yep.

I think that is probably going to be more the use case than, I mean, again, even with Firefly image, like, I think that most people use it in Gen fill, people don't really go to adobe.com, backslash firefly and you know, image model and then generate stuff there. You know, they do it, they do it in Photoshop. So I think that's really, that's really where it's going to live. So I don't even necessarily know if, like Adobe necessarily really belongs on this list because it's not necessarily. I don't really think of it as like a full platform generator as much as I do an Extension of Photoshop and Premiere.

Yeah, gotcha.

Yeah.

Well, and then the last one we've got on the list here is the make a video from Meta, which nobody has access to yet. So is like, yeah, I don't know if we will or not, but maybe they'll make it open source or, you know, Facebook's version of open source. Every time I say that Meta is open source, I get a million comments from people going, it's not open source. Stop calling it open source.

Oh, really?

Yeah. But, you know, meta's level of open source. Maybe we'll. Maybe they'll open source. Open source. Ish. This one.

I'm curious, like, with the. Now that we're sort of getting this, like, second tier of like. Of like. Because, I mean, I think that this is maybe like in my head, since I've been following this for so long, I consider this era that we're moving into as like, AI video 3.0. And like, definitely with Sora now released and with Vu released, I mean, it's kind of. It's. It's kind of forcing. Met his hand a little bit to do something with this. Right? I mean, I don't know if Zuck really cares, but, you know, but I.

Mean, you know, just comparing it to what else is out there. This looks like the last generation of AI video, right? Like, it doesn't look like the current gen that we're seeing now, but this.

Was the one that had all of the sound effects and you can do. You know, everything generates as one. Right?

Feels like a. Feels like a Pico competitor. It feels like Pica is going to go more in a social direction and then Facebook's already there, so they're kind of like, if they approve this, they'll end up kind of causing problems for Pika possibly, like, if you made like a cool social product, you know, with it.

My other question too is in. Yeah, in this, given that it is Meta, like, where does this live? Is this. Are they thinking about this for. Is this a Facebook thing? Is this an Instagram thing? Like, where does this live in the Meta ecosystem? Like, like, what is it? Yeah, what's it? What?

Why?

I still feel like. I've been saying for a while, I still feel like there's some new social product with AI that doesn't exist yet. I don't know what that is. It could be they're probably. Probably multiple ones, but, like, Instagram started because of filters. Well, maybe there's something new now that you could create with with these new AI technologies. Like Whether it's like, video and you take a photo of yourself and now you made a really cool viral video or something. You know, something really funny.

Well, and I know that's the other thing that everybody's really hot on right now is, like, trying to figure out or not everyone. There's a lot of companies that are. That are very much trying to become the Netflix of AI video. And like, I. I'm not. I don't believe that needs to happen yet. I think that we need to have AI video on Netflix. I think I would like to see an AI generated. You know, I produce, produced, like, thing, like something like, did you guys watch Scavengers Reign on hbo? So cool. So good. It was that animated series about the.

There's no HBO in Japan, which is such a bummer, really. There's no HBO in Japan.

Called a vpn.

Vpn. I was going to say.

Yeah, edit this out.

Vpn. And the iron A.

What sucks is, though, but. But I watch content. I watch stuff with my wife. Right. And so you want the subtitles. So it's like, unless I'm like, you know. Yeah, like, like, like we're. We're finishing up watching Vikings right now, which was kind of funny for me because we're doing all this Viking AI video stuff here. We're finishing up that series right now. And I was like, well, I like Game. I'd actually never watched the Vikings before, and it's pretty good. I was like, I like Game of Thrones better, except for the last season. And so I'm wanting her to watch Game of Thrones. You're like, she's a huge fantasy nerd like me, and she's never watched Game of Thrones. And it's just like, I had to buy the entire show. Like, it's quite expensive, like, actually buying it.

So right now, this is. This is our current rankings of the tools. This is where I feel like they all landed. Is anything need to be adjusted? What do you guys.

Zuckerberg is never coming on our podcast.

Oh, he'll be fine. Well, I think going back to, like, the only reason I put Meta down in the D tier is A, we don't have access to it. B, it still kind of looks like the last gen AI, like, Meta. Tomorrow could drop something that looks just as good as Veil and be like, we've been working on this behind the scenes and all of a sudden they're an A or an S tier. Right.

I think they'll have some cool stuff coming out. Same as. I think you know, Grok will have some cool AI video stuff at some point too. And, you know.

Yeah, yeah. So, I mean, obviously none of this is a knock on any of the companies. We're just kind of like ranking them based on where we feel their current generation of technology is. And any one of them, like, is an update away from, you know, bumping up a tier or two. And also, like, like Tim mentioned, I don't really feel like we know what meta's gonna do with AI video yet. Right. Like, they are trying to shoehorn, like the AI image generators, like, into their chat apps and stuff, but I don't think anybody's still really using them.

I don't think that's gonna work. Just adding it on stuff. It's like the same problem Adobe has. You just add on AI to some existing thing. Like. No, you need entirely new things for this new world.

That's.

Now there's. There's so many new possibilities. Why do you want to be tied to these old systems? But they have to, because that's what that's like. They're, you know, cash cows. Right.

Well, and I think, I think that as we move into 2025 too, like, there's going to be a big shift up amongst all of these and we're going to see if we come back and do this again next year, you're going to have. We're going to have like five or six other companies that we've never even heard of now on this list. The list is going to be all over the place. We're going to see a lot more of that 3D element with what we were talking about before with World Labs and with what kind of, in a lot of ways, what Genie 2 is doing. Those, like, world builders, like. And I know that Mid Journey has been talking because Mid Journey Video will be coming along.

That's what I was about to say.

Yeah.

They've been wanting to do. Yeah, yeah.

So Mid Journey is also working on a video model too that they keep on saying they're going to release.

They haven't shown it because I'm sure, yeah, there's so many new, like, great video models. Like, how do they release something without it being spectacular? Right.

Like, actually, and I was just in the office hours a couple of weeks ago. I think they, they were like. And all power. I'll bless them for it too, as well as the team was just like, we could release it. It's the holidays. We'd like to spend time with our families. Like, so that's like legitimately why they're not, you know, because like if they release it, they're going to be hit with all kinds of like questions and they, and the server's going to go down this and that and that. Like just. Yeah, let them, let them go home and hang out with their families and.

Then, and let's be honest, right now is not the time to be releasing something with what OpenAI and Google are doing. Right. They're just going to get overshadowed by those two companies right now who rock.

Em sock em robo at it right now. Like so, yeah.

Meanwhile everything's speeding up.

Oh yeah.

What's your prediction on January? Do you think we're going to see a massive slowdown or do you think we're going to ramp up again?

So I, I'm kind of waiting to see what OpenAI releases over the next two days.

Yeah, right.

I'm sort of giving away when we're recording this episode compared to when it's getting released. But there's two more days of OpenAI about to release updates. I think if they don't release tool use as one of the next two updates, I think that's coming in January. They've already hinted that that's probably going to come in January. So I think, I think we'll see the tool use come from OpenAI and then, you know, knowing what I know about Google and spending some time there, they're really working on the sort of like agentic tool use kind of stuff as well. So I think that's going to be the big theme of 2025 overall is these tools that can use tools on your behalf.

Yeah, I'm excited about that.

I, I don't think we're going to see like a new AI video model for instance, in January, unless it's from like mid journey and they've already got it ready to go, you know.

Yeah, I think we'll see the rate of progress improve. Like if everyone, like probably everyone's gonna be copying like what OpenAI is doing in terms of like their O1 model and from my understanding, like, you know, since it's not entirely just reliant on the training of, you know, with massive amounts of data. That's one part of it. But that's not the only way it improves. We'll probably see these models just like keep getting better, like over and over and over. I mean we might, we might get to a point where like every month there's like a. It got better 20% this month.

Right.

And I think AI video will end up in a similar scenario quite soon, if not, if it's not already there.

You know, it's funny because I talked to Caleb over at Curious Refuge about this a bunch of times. He and I chat back and forth of like, you know, as the video models get better and better and better and like, you know, we are always reporting and doing tutorials on all of it and like, what's the, what's the point when it's like, it's like, okay, it's perfect now. Like, what is. Like, what would we make then? And like, the fact is, like, it'll never, it'll never be there. Like, it's just, it's just like there are still people doing, you know, camera tutorials and camera like, you know, reviews and, and all of that. There's always, you know, it's like last year's Canon is terrible. This year's Canon is better. Nikon is better than that. You know, there's always just going to turn into that. So.

Yeah, yeah, I agree. I think it's gonna get better. I think we're get higher and higher resolution. I think there's always going to be room to focus on the finer details within images. You know, we're going to get more controllability out of the images. Right. Where you have a little more control over the lighting, over the, the style of it. We're already seeing that with, I mean, basically Mid Journey's Got Lauras available now. They don't call it that, but they've. Yeah, the mood boards is essentially that.

So I think there's always gonna be places to improve, but it's gonna be more and more nuanced as time goes on.

And I think one of the things that I kind of wanna start maybe like preaching a little bit more on the channel is like, is start moving into some of the tools that help you develop story. Because ultimately as fidelity gets better, as details get better, as control gets better, what are we doing with this stuff? Like, are we telling a good story? Like, is it something that's gonna make us laugh, cry, get excited, whatever? I mean, that ultimately is the thing. Like none of us watched Star wars when we were kids and, and you know, we were, we all love the special effects, but that's not what drew us in, you know?

Right, right.

For sure, for sure.

And I've actually really started to see some impressive AI generated short films. I saw one that somebody made recently. I'm sure you've come across it, Tim. In fact, it might even be your X account that I came across it on. But there was recently, like, a little Batman short film that came out.

Oh, yeah, yeah, yeah.

Really, really good. That got copyright really good on it. Oh, did.

I went back looking for it and they got knocked down for being. Yeah, it was. Warner had knocked it down. So.

Yeah, I mean, I'm not surprised by that. They did basically just, you know, steal somebody else's IP and make a short film out of it. That was really good.

It was actually. Yeah, it was really solid. Great sound design throughout that whole thing as well. They really captured the tone. There was an interesting trick that those guys pulled where they used that first frame feature, but they were taking screenshots from the Robert Pattinson, you know, Batman movie and then basically running their own generations to kind of create iterations of that scene. It's really smart in all honesty, but I can see exactly why Warner was like, no, no, no. That is gone now.

They'll release something like that on their own in the future. Right? Like here.

Yeah.

Make your own custom Batman story.

I'm sure, I'm sure. But I mean, like, just that idea alone opens up the door for so many, like, fan films and that kind of stuff. I actually, truthfully, I think it's kind of stupid that Warner knocked it down. I really think that it's like, if anything, that's just like free publicity for your Batman movie. So I don't, you know, I'm not a corporate lawyer, though.

So, again, it's. The point is people are figuring out how to use these tools to make. Like, knowing the limitations is just creating bumpers that they know they have to stick between. And, you know, with those bumpers, they're figuring out how to make stuff look really good and sound really good. And I think we're just going to see more and more of that. You know, ideally people should create their own IP instead of take it.

I agree with that.

Marvel or companies like that, because that kind of stuff's probably going to get slapped down every day of the week. We saw it with the south park generation too, where there was a whole AI generated South park episode, which was actually a decent episode. I watched the whole thing. It was like, that's not bad. But that also got slapped down. Right. That got pulled off the Internet because that's somebody else's iPad.

Yeah, the simulation guys are sort of moving out of that south park thing and now they're starting to do kind of like their own branded IPs. And the last one, I think that it kind of looked a little bit more like that Star Trek lower Deck show, it's not. But it kind of has that vibe to it. It's like a spaceship thing and kind of an animated series. But that's. Yeah, I agree with you. I think that, like, just don't make fan fiction for stuff. Go make your own fiction. Like, you know, like, it can be very heavily based on, you know, the thing that you want to make. It's not the Punisher, it's the.

It's the punishment or, you know, like. Or whatever you want to do. The Vindicator. Yeah, exactly. Yeah. So, you know, just make that instead. So at least you own it.

Well, cool, Tim, this has been amazing. You know, obviously this episode isn't two hours long, but for anybody who's listening right now, we've actually been recording for two hours. That's how Nerd out. We've gone on some of these video tools. But before we do, like, shout out your platforms. I know you've got an awesome YouTube channel you post on X. Where should people go check you out?

That's really about it right now. I think next year I've got to start. We'll probably pick your brain a little bit on like, on like doing a website and probably a newsletter at some point or another. Actually pick both.

Tell people the channel name and the. And your.

Oh, sorry. Yeah. So theoretically, media on YouTube is probably the easiest way. And then from there you can find links.

Absolutely. Absolutely. Well, thanks again, Tim. Thanks for everybody who's tuning in, make sure that if you're not already subscribed, you subscribe on YouTube. That's where you're going to get all the cool visuals. You're going to be able to see our breakdown and examples of these tools. You're going to be able to see the tier list that we made.

See the handsome purposes of Matt.

See our handsome faces of Tim. Nathan.

I was hoping you were going to say Nathan too. You're going to say Matt.

Well, this is awkward.

But again, if you do prefer audio, we are available wherever you listen to podcasts. So thanks again, Tim, and thanks everybody for tuning in.