SeaDance 2.0: el salto en video IA con audio nativo (Full Transcript)

[00:00:00] Speaker 1: SeaDance 2.0 is here and it fixes basically every problem with AI video generation. Character consistency, realistic physics, multi-shot sequences and native audio sync, all in one model. Let me show you everything that's new and you can click the first link in the description down below to try it out for yourself. If you've used SeaDance 1.5 you know the pain. Characters would morph between shots, the motion felt a little bit robotic and you had zero audio control. SeaDance 2.0 addresses all of that and in this video we're breaking down every major improvement and what's changed from 1.5 to 2.0. The first thing is the multimodal input system. You can now include up to 12 files per generation. Nine images, three videos, three audio files plus the text prompt. On top of that it also has the tagging system so you can mention specific references and assign roles to the different assets. So for example you could have at image one for your character, you could have at video one for the camera motion and then at audio one for the voice or the audio that you want to use in the generation. And so here you could go and use text-to-speech and 11 creative, go and generate some music and bring it all together to generate your video. SeaDance 2.0 also has multi-shot storyboarding which allows you to generate connected sequences and not just single clips which is a major improvement and very useful for AI creators and filmmakers. With SeaDance 1.5 there was a ton of issues regarding consistency whether that was for faces or clothing but 2.0 locks the character's consistency across the entire sequence that you generate. The quality of the motion in the video has also significantly improved going from something that could occasionally feel robotic and quite generic to movement that has realistic physics with proper gravity, momentum and collision. So there's a real-world understanding within your video when you generate it. And with SeaDance 1.5 and other video models there was often issues with textures and flickering where things would randomly shift between frames but in SeaDance 2.0 I am yet to see any issues with weird flickering between frames. Now most AI video generation tools generate videos where sometimes it feels like the audio doesn't quite match but SeaDance 2.0 generates audio and video at the same time which means that sound effects actually land when things happen on the screen. If a character is walking the footsteps match or if the door slams the audio hits right on the frame. And SeaDance 2.0 also has beat matching with its generations and this is where it gets pretty interesting for music content because you can upload a track and the model reads the rhythm and generates visuals that sync to the beats. So if you're making music videos the dancers movements actually land on the kick and the snare and transitions sync to the drop of your music which used to take hours in manual editing and would often be offbeat in other AI video generations. So here you could go and generate the perfect track with 11 music and then bring that in to guide your generations with SeaDance 2.0. With SeaDance 2.0 you can generate video in over eight languages including English, Mandarin, Spanish, French, German, Japanese and Korean and because it's generated alongside the video the lip sync is a lot tighter. But what you could also do if the language wasn't available is you could generate your voiceover with 11 Creative Text-to-Speech, feed that in as your audio reference, maybe add a music track for the rhythm too and SeaDance 2.0 will sync the visuals and the audio to match the lips and the timing of the generation. And because you can use so many different assets for your generations you can get some really unique outputs but you also have so much control over what the final video looks like because you control the sound whether that's the voiceover and the music you can even add sound effects and then you can add again up to nine images as references and three videos for the motion. And it's not just about generating video from scratch you can take a video that you already have and regenerate specific parts of it while keeping everything else intact. So if you want to change elements within a scene or swap out the character entirely SeaDance 2.0 handles that while keeping the rest of the video consistent to what it was before. And if you want to try it out you can click the first link in the description down below and try SeaDance 2.0 inside Eleven Creative. Try out the multi-modal output, the multi-shot storyboarding, consistent characters, realistic physics and native audio sync and let us know what you think in the comments because we would love to hear your thoughts. And if you want to see more breakdowns about future model releases hit the like button and don't forget to subscribe. Thanks for watching.