[00:01] released what people call the death of Hollywood, Seedance 2.0. And with this film that actually looks professional. really makes it the best AI video generator and how it destroyed all the
[00:13] popular models like Sora 2, Google V O 3, and Kling. Seedance 2.0 is now live yourself and not fall behind this fast-moving technology, I'll leave a look at the five capabilities that completely changed this new AI model:
[00:28] character consistency, realistic physics and people, camera movements and shocked me the most, but before we get to that, let's see the character consistency. Almost every single AI model can create videos with two
[00:40] methods: text to video and image to video. And this is no longer impressive. that changed everything. It's called video to video. And here's where it gets wild. You can upload up to three videos, six images, and an audio file all at
[00:53] once and use them as references for the camera movements, video style, action, more control than you ever thought was possible, but it also gives you one of the biggest advantages out there. Most AI models, including Seedance 2.0, are
[01:06] But with the video reference feature, you can take the last 3 seconds of a generated clip and use it as a reference for the next clip. So in theory, you can make 3-minute videos or 2-hour-long films all in one tool. That's exactly
[01:19] films all in one tool. That's exactly how this 56-second clip was made. the other models because the fire and ice textures are so realistic. The body
[01:32] movements feel super natural, but what's really amazing is that the characters stay throughout the entire video, which has 56 seconds. This is a level that no scenes transition so naturally and the shots look great. There's also no
[01:44] morphing at all. The overall video is really, really solid. I actually have amazing how before Seedance 2.0, we were judging AI clips based on morphs or watching a movie and trying to find small mistakes in it. And this is a
[01:58] generations stitched together. Now, I don't know about you, but I don't even the next starts. Everything is just flawless, but it could be just this video that looks so good. So, let's see another clip that uses multiple
[02:11] characters to see if they stay consistent or if there are any minor consistent or if there are any minor mistakes.
[02:28] bodies. And this slow motion intensifies the entire moment, making it more everything else looks so realistic. The characters, the background, and the movements all feel so real. And I love how they combine the punches with the
[02:41] through. I'm really curious what inputs are behind this generation because it looks so clean and detailed. Also, it's a 25-second clip, but it manages to keep the text from the walls consistent all the way through. SeaDance 2.0 is simply
[02:54] character consistency across different useless in a real movie if the model is not good at replicating realistic physics. And as you know, Sora 2 was the king of physics up until now. So, let's
[03:07] king of physics up until now. So, let's see if SeaDance can take its place.
[03:24] motivational sports ad. Every movement, every punch, and every hand entry into the water feels so realistic. Sora 2 was good, but this right here is a new level shadow for the woman running. Also, the lip sync is perfect. There's no morphing
[03:37] really impressive. The scene cuts are very wholesome, and even the color impressive. But this seems to be something common among all the SeaDance generations. Just take a look at this Formula 1 racing.
[03:53] imitates the car suspensions. Every time the car makes a turn, the wheels from real life. I actually feel like I'm watching a Formula 1 car race on TV. The The camera angles, like the grip shot
[04:06] the sound matches reality. And if you pay attention, you can even see the rain believe it can't get better than this. But what's actually unique is the text on the moving cars. There's no morphing at all, and there's no other popular AI
[04:19] model that can keep this consistency for 15 seconds. Sea Dance is now definitely the new king of realistic physics. But what about realistic human interactions? unusable. Kling is the well-known model for creating UGC content, so let's see
[04:32] it to a stress test, I want to see how it generates UGC content where a girl puts moisturizer on her face. The reason I'm doing this is because it's one of you've tried it before, you know how many credits you can spend without
[04:46] problem is in the small details. There's always something that makes your brain realize it's AI. So let's see if Sea Dance can take this test.
[05:03] actually makes the content seem so real. The acting is great, and the movement is test, and it actually did more than that. It kept the typical style of UGC, The brand name stays consistent on the product, and the lip sync is almost
[05:18] they can do the same even if there's a lot of action going on. Here's a POV fight that highlights it perfectly. >> [screaming]
[05:39] And I'm not sure how Sea Dance does it, but it keeps everything smooth and are on point, but that's not the main reason Hollywood is panicking. Up until now, AI models could imitate physics, human interactions, and keep them
[05:53] was almost impossible was cinematic multi-shot sequences. Now, SeeDMs 2.0 prompt. So, the scenes that took millions of dollars to create in person from their room. And here's the proof.
[06:22] missed was the number of swords. In the beginning, the man on the left has two has one. But apart from this, all the details are in place. Just take a look smoothly that big lamp falls. I love how for every couple of seconds, the camera
[06:36] of multi-shots, but the transition between them is so smooth that it seems natural. I once tried to do a scene like this with Google VO3, and I remember the has nailed everything, including the VFX, which are really on point. This
[06:50] are more generations stitched into one. But again, the transitions are so well one ends and another begins. This scene studio. And besides that, you'd need an entire film crew and actors. But the
[07:04] couldn't even be recreated in reality because the characters are not humans.
[07:22] a dog mid-air? This is the strangest scene I've seen today, but it's still with complex camera movement. It's more of a combination between TikTok and took to create this, but I'd love to see the input behind it and create one
[07:37] the speed ramp-ups and the camera movements makes the whole scene so dynamic. And now, there's one final capability that C Dance 2.0 uses to talking about precise motion. So, let's watch this clip and see what I mean by
[07:51] watch this clip and see what I mean by that. and I can't see any morphing in the whole clip. So, this is definitely models. And honestly, I'm very impressed. I think this is my favorite
[08:04] one from the entire list. Just imagine combining all of these capabilities should really be concerned about how good C Dance is. There's a user online who claims he gave this new AI model just a picture and got this insane
[08:17] just a picture and got this insane motion design.
[08:33] that, it's all solid. And if you think this was all done by C Dance from just a photo, you start to think that this AI video generator has a brain. And that's most AI video models, this one starts a
[08:46] get enough context from the user. This is what makes it so good at hyperrealism, consistency, and precise motion. So, C Dance will really change want to test it yourself, I'll leave a link to Higgsfield down in the
[08:59] out today. Thanks for watching, and I'll see you in the next one.