Seedance 2.0: Глобальний реліз, обмеження та альтернативи для генерації відео

Well, Seance 2.0 has finally made landfall in the United States. So, officially, it is now global. So, today we're going to take a quick look at it. Namely, and totally non-sponsored, where I think you might be getting the best bang for your buck on it, but we're also going to talk about the restrictions. And I've got some good news on that front and you know, while some workarounds in the meantime. Plus, we're going to check back in with Pix's updated realtime video model. This is one that you can play with right now. And actually, it's getting pretty cool. Kicking off, yes, Seance 2.0 has finally landed on the shores of the good old USA, meaning that it is now truly globally released. And I don't have to keep making those Seance releases kind of thumbnails anymore. Now, alongside this release, it does seem like, you know, the imagetovideo uh and Omniref face filtering has ramped up a notch, although you still can get away with some stuff as we're going to see in a moment. But yes, the model is now finally available over on the mothership. Dreamina, uh, Fall has it and a number of other API platforms. So, thankfully, we are done with those janky weird sites that claim to have Cance 2.0, but were really either serving you up with Cling 3.0 outputs or just, you know, taking your money. But again, alongside this release is like a clamping on the various restrictions. It's never easy with Cance, is it? Uh, I do have some news on that coming up in just a minute as well. But in the meantime, and overall, I'm not picking one platform over the other. You guys know that I cover them all. And again, this is totally unsponsored. As a side note, there is a sponsor spot later on in this video, and it is pretty good. So, now that it essentially is everywhere, where is the best bang for your Seance buck? Currently, I'm going to put my money down on Runway. Now, Runway has already released Cedance 2.0 know on the platform. Although they are lagging just a bit on the US release. That said, I do have it on very good authority that this will be changing very soon. Chances are actually by the time you're watching this video, uh it will have already been released. In the meantime, I'm not not saying that you can't use a VPN to very easily access it. But, you know, by applying the scientific method, I am not not saying that it doesn't work. Now, again, the reason that I'm picking Runway for this is that we do know that Cedance 2.0 0 is a very expensive model and we also know that runway does offer an unlimited plan with seedance 2.0 actually included. Currently it does say accept US but again that is going to change pretty soon. So yeah I mean essentially 76 bucks a month or $95 a month you have unlimited seedance 2.0 generations plus you know everything else that runway offers. Now I do have to point out that on the runway unlimited plan your generations are capped to two at a time. uh when you're using that part. You have credits to burn initially. Um but you know after that you are also on the slow boat. So uh you know generations might take a little bit longer in the queue but it also doesn't cost you any credits. Obviously I do not have like the pricing plans of every other platform that's going to be offering Cance 2.0. So this is subject to change. That said I do have to say that unlimited is going to be a pretty hard bar to beat. Anyhow moving on to the current restrictions and limitations. Yeah, these can be a bit on the tricky side right now. And I do want to note that again, while we're testing essentially all of this on runway, I do not believe that any kind of the content restrictions that are happening here are on the runway side. This is all, you know, bite dance and C dance 2.0. So, oddly, well, starting off on the positive, flamethrower girl, I mean, she seems to have no problem whatsoever. In fact, testing her out against another character, Lyra, who's been sitting around since the uh like Viking dark fantasy, the bridge era. Um, yeah. I mean, running these two together, fighting each other, no problem whatsoever. That said, a test with Tom and Sunny from Planet Hell, which we have run a few times on the channel in the past. Uh, and yes, I am still pottering our way on that short film. that comes out completely blocked. Funny enough, I did just recently pull this clip off in Cling 3.0 using their Omni reference model. Uh, if you want to see that, plus a bunch of other Clling clips. Um, well, funny story, Clling actually had me do a guest host spot on their channel for a tutorial. So, uh, I'll have that linked down below. I actually didn't even know it went up. It was my buddy Chris Branch that uh, texted me and let me know. Thanks, Chris. So, moving back over to the runway and sea dance side of things. Uh, inversely taking that same prompt and instead of using Sunny and Tom, I just switched them out for Flamethrower Girl and Lyra. And >> this girl, >> you're never going to make it, new girl. >> Yeah, it totally works. It doesn't make any sense, but I mean, it totally works. Although Flamethrower Girl is really mean in that clip, isn't she? So, obviously, it can be done. I'm just trying to figure out exactly where the triggers lie. Uh, I mean, it does constantly block Renfield as being a real person, which I'm sure he has a lot to say about. >> Why does the computer be thinking I be a real person? >> But one place we are not running into any trouble whatsoever is with animated outputs. Uh, this is another character that we used way back in, I want to say, the Kyber days. and running her through a quick animated style prompt, we end up with this, which I this looks pretty great. So, that is one avenue to explore while we're waiting for these restrictions to soften up a bit. Now, another trick you can pull off is to begin by running a text to video generation with say a location, a location that we have used a number of times on the channel in the past, but no characters. Um, so we can run this. >> Mr. Malloy, I hear you're a detective, and I'm in need of one. >> Detective? No, Snoop maybe. What's the case? >> And then by utilizing Cance's capabilities to continue the scene, we can well continue the scene. >> It's a murder >> and I think you're going to be interested. >> Yeah. Yeah. Who's getting iced? >> You are. >> Theoretically, hey, name of the channel. you should be able to chain out consistent characters for like a scene and maybe even transpose them to another location, although I have not tried that yet. And I do want to say that one important part here is going to be experimenting with prompts. Uh, for example, in our usual FBI agent having coffee in a Pacific Northwest diner, I mean, very Twin Peaks coded. This is a damn fine cup of coffee. But how's the pie? It's really good. It's an old recipe from the old lady who got it from the owl. >> So, I think that's a pretty great version of that example. Although, you may have caught at the very end there that it was the old lady, not the log lady. I will say that did take a little bit of experimentation. The original prompt is uh cinematic surreal crime drama. Again, very heavily Twin Peaks coded. Uh FBI agent in a Pacific Northwest diner drinking a cup of coffee. Um, we even have it's a damn fine cup of coffee. Technically, it should be damn fine cup of Joe, but um, you know, obviously, uh, we got push back on that. So, I didn't know if it was the prompt holistically that it was pushing back on, but after some experimenting, no, as it turns out, it was just log lady. So, change to old lady passes fine. And listen, I know this is going to be frustrating. How do I know? Well, because this was my morning. So, um, yeah. No, I mean, I totally get it. Uh on the plus side, you know, you're not charged for anything that pushes back. So you're not uh you know, losing any credits here or anything. And if you have the unlimited plan, I mean, well, there's no credits to lose. Now, I can say, at least I think I can say, well, I'm going to say it anyway, so it doesn't really matter, that uh Dreamina has begun allowing those in its creative partner program, full disclaimer, I am one of them, uh to allow the use of realistic faces. So I've had access to the model when it was completely unhinged. uh you know the slightly hinged version that I did uh Dragon Blue with and then you know now obviously the more moderated version and now back to I think where it is is is we're never getting the full truly unhinged one again. But essentially the version that I used for uh the flamethrower girl short. Yeah, that I think will be the standard model provided everyone behaves and we don't have a news flare up of like Tom Cruz and Brad Pitt fighting again. I think what's generally implied is that since it's released to all creative partners, it's almost like a like a beta feature and will be released to everyone probably within a week or two. In the meantime, I'll keep chisling for workarounds on the restrictions on Cance 2.0. Uh I'll compile what I find in the newsletter, which you can subscribe to that is linked down below. But for now, we're going to go check back in on Pix versus Realtime video model. This is actually pretty cool. Uh, but first a quick word where I get to play around with Lyria 3. Um, I think actually this is this is a pretty fun uh sponsored segment. You're definitely going to want to check this out. So, Art List, who I think most of you probably know and you I have previously covered with their AI toolkit, well, they are now rolling out music generation. And this one is pretty cool because the model under the hood is Google's LIA 3. And what's interesting about LIIA 3 is that while I've only had a little bit of time to play with it, some music pro friends of mine have been saying it's become their go-to generator. I mean, that said, everyone with ears has an opinion on which AI music generator is best. So, let's dive in and check it out on Art List. And what's cool is that they've actually built some extra tools around it. The way it works is pretty simple. We just make sure that obviously we're on the music tab here. Prompt box is here. We have an enhanced prompt over here. And then uh with Liia 3 Pro uh if you hadn't heard the news, we can actually generate up to 3 minutes. We also have options for mood, theme, and tempo and genre as well. And to note with any of these, like you don't have you're not stuck in a box uh with any of these selections. You can prompt uh outside of any of the things listed there. And obviously we can prompt for instrumental, custom lyrics, or auto lyrics. Um, so just kicking it off with an actually an older test which is an epic metal song about a Viking searching for the perfect cup of coffee. This gets pretty metal. The dark sea rises, the quest is done. The black sun for the morning sun for the power, the caffeinated MIND. WAKE THE WOLF AND play the light. Now there's a song that'll get you up in the morning. One thing that really caught my eye on Art List is image to music. So yeah, you can load up to seven different images. Uh in this case, and admittedly this does get a bit silly, uh this was an image of a pelican riding a bike. Uh you know, the old SVG test. Actually, I think that this was this was actually generated in Nano Banana 2 for our video on that. So kind of keep it in the Google family. And the prompt here was really just as simple as create a song based off this image. Uh, and we ended up with >> blue paint, shiny frame. Check the flags and know the name on looking cool, breaking every bird and rule. He's got the snacks and he's got the style. Cycling down the golden mile. Pedal power, pedal pride. Take a look at the pelican ride. I mean, that is pretty impressive. Goofy, yes, but I mean it got a lot of those image details in there. Rounding out, one thing that I definitely wanted to point out is that there is now artless sound as well. So, if you have this option toggled on, this is essentially a post-processing layer that polishes up the output to well, studio quality production standards. So, uh let's give that a listen with, uh, create a hip-hop song about how nothing rhymes with orange. If you've been following the channel for a while, you know things do rhyme with orange. Just the cupboard is bare, but I'm sniffing for the scent. There's a flicker in the air. Door hinge more binge. Keep the flow in the gym. I'm a Norse king with the storm on the rim. It's a hellar of war when I'm leaning on the port swing. Everything is gorgeous when you're sipping on the orange. I mean, that is pretty good. Door hinge is something that we often hear in these music generators. But I got to admit, like Norse King to port swing, I I that's the first that I've come across that one. It probably makes our coffee swilling uh Viking from earlier pretty happy. And using the artless sound option, well, I mean, you could definitely hear a difference. there's like no more playing around with limiters and multiband compressors afterwards. Um, so yeah, another notch in the growing arsenal of creative tools available over on Art List. And the nice thing is AI music is included with all of their AI plans, starter, professional, and max. So, uh, if you want to check it out, the link is down below. Thanks as always to our friends at Art List for sponsoring today's video. See, that was a fun segment. I actually still have that stupid Pelican song stuck in my head. Uh, moving on. uh Pixverse as uh they of course released their C1 uh video model just two days ago. I still haven't gotten a ton of time to look at that one but you know overall it is looking pretty good. Um but in addition to that they actually released an update to their real-time video model which I did take a look at a little while back to note that what we're looking at here is actually not you know the actual real-time video model. Uh, this was, you know, a sizzle compilation that they put together, probably utilizing the new C1 model. I'm not entirely sure about that. Um, but this new updated version allows you to essentially input your own avatars and characters into uh the real-time video and it's, you know, it's definitely gotten a little more stable. Still pretty wonky though. Now, you should be able to access the real-time generator uh right here under the try now button here. that'll kick you over to um the well the real-time video model. There's a number of community-made scenarios that you can jump right into and start playing around with, but I mean obviously you're going to you're going to want to have fun with your own scenario. More than that, you can actually now create an avatar to play around with in real time. Um you know, obviously I'm going to put flamethrower girl in there. Uh and then from there, all we have to do is create a world. I've been pretty excited to see the back rooms, which if you're not aware of is kind of like liinal horror thing that got it start right here on YouTube, uh, and will now be an A24 movie. Um, so I created sort of a, you know, back rooms inspired kind of thing, like literally called the back rooms. Uh, so, um, text description down here, uh, a image, uh, that is very much inspired by the back rooms. Uh, and then we have three choices here of POV, ambient, or dramatic. Uh, I have found that ambient is the way to go with this one. if you go dramatic, uh, things, well, things just start getting weird. Um, so let's go ahead and explore this model. I'm going to run this more or less in real time just so that you guys kind of get a vibe for it because yeah, I mean, look, it's wonky, but it's also super fun. And we're off. I'm going to run this relatively in real time. I'm going to mute this audio here. I think it still kind of comes in. Yeah, I mean, this is all being generated in real time. It does have a little bit of like a cartoony vibe to it, but it's definitely gotten a lot better since the original um uh version of, you know, this real time bottle. So, she's running around like I guess I guess she's doing a dolly shot with a DSLR. Um so, let's uh take a picture of one of the desks. See what she does here. Um yeah, I mean, look, it's still going to be wonky and weird. actually kind of what I love about it is that um you know I mean this is once again just old is that a is that a hot pocket that's there. Um see take a picture of the hot pocket that's on the table or is she just going to go over the vending? Oh, she tripped on a who's who left coffee there. See the back rooms are evil. Um so the coffee just You got to watch out for that coffee. What happened in the hot pocket? Um pick up the cup of coffee and throw it in the garbage can please. Um, before somebody else gets hurt here. Root board. It's wonky. It's weird, but it's also kind of Oh, and our our camera gets turned into a coffee cup. Yeah. Again, uh, it's going to be morphine strange. Why? What are you doing? This is incredible. Um, all right. Put down the camera and pick up a flamethrower. Now, I have often said that when everything in AI video is absolutely perfected, and I mean, there's certainly an argument that we're getting very close uh given Cance 2.0 that I really am going to miss this era of weird, warpy, odd, nonsensical AI video. Um, you know, that said, clearly real time does still have a ways to go with this and and is kind of the home of uh, you know, surreal AI video. Um that said, you know, given the rate of progress, a real-time output like this, I mean, eventually it's going to look like Cance 2.0. Now, whether that happens a year from now, you know, 6 months from now, or like a week from now, I don't know. Um, probably not a week from now, but also I've given up estimating times because you get progress always surprises me. So, yeah, head over to Pix. I don't think this costs you any credit, so just, you know, log in uh basically hit real time and and, you know, have some fun. So, I guess that's it for today. I am actually traveling next week, but uh I might do a report from the hotel room. I do hate doing that, but uh we'll see. But in the meantime, I'm going to try to have something prepped for you guys at least on Monday. Um so, yeah. In the meantime, I thank you for watching. My name is Tim.