These 3 AI Video Tools Will Be Bigger Than ChatGPT

Today, we're talking about some upcoming AI  apps that don't even seem real. Once they come   out, these apps are completely gonna change  the media landscape. And it's really useful   to know about them in advance so you can start  planning for what is to come. The technologies   and research we will talk about today will soon  be a bigger part of your life than even the most   viral TikTok trend. So, get ready to discover  some cutting-edge AI research and applications   that are said to change the game in a big way.  And remember, the only thing scarier than AI   taking over the world is you not knowing about it.  Okay, so let's start by talking about Flawless AI.

Now, you might have heard about this as these  videos have been circulating around social media   since a few months, and particularly  this clip has been going viral. So,   let's look at it and then talk about some of  the implications. So, as you can see here, this analyzes faces,   and it is able to change facial expressions and  especially the mouth, which allows for flawless   translation after the fact. Now, if you've been following the  developments in the text-to-speech department,   that technology is getting scarily good. One might  say indistinguishable from an actual speaker. That   is the last time you steal my voice. If you add  the fact that large language models like ChatGPT   are really good at translating text, even better  than classic translators because they're excellent   at assuming context. The last missing link in  order for us to translate any movie, video,   even podcast to every single language on Earth  is exactly this. Because large language models   like ChatGPT are going to be able to translate  the transcript from English to, let's say,   Japanese, then you'll have a text-to-speech AI  that, for example, was trained on Tom Cruise's   voice read that translated transcript. And last  but not least, this software is going to adjust   the facial expressions and the mouth movements to  match the language. And I have to point out that   there is a dark side to this, right? That means  that the fakes are going to be undistinguishable   from real videos very, very soon.   Scary. But the brights side is that people will get to share their culture and their   knowledge all across the world. Imagine listening  to the most popular Chinese AI podcast in perfect   English. I think that's extremely valuable, and if  you produce any kind of content, this is something   you have to keep in mind because soon you're  going to be able to distribute to the entire   world. As with most of these technologies, when  exactly it's coming hasn't been announced yet.

Next up, let's talk about Runway with  their generative video technology that,   just from the demos, looks completely insane.  Gen 1 is a video-generation AI system that   can efficiently generate video in any style all  while retaining quality and flexibility. Gen 1 is   able to realistically and consistently apply the  composition and style of an image or text prompt   to the target video, allowing you to generate  new video content using an existing video.   We call this approach 'video to video,' and we're  incredibly excited to share a few early use cases:   stylization. And look at this, this allows you  to bring some of the capabilities we already   know from ChatGPT, like applying styles to text,  to video. This allows for a whole new level of   creativity. Look, with a bunch of books, you  can recreate a drone shot over a busy city.   Mask mode isolate subjects in your video  and modify them with simple text prompts,   and customization mode unleash the full power  of Gen 1 by customizing the model for even   higher fidelity results. Gen 1 represents  a pivotal step forward in generative AI,   one that brings us meaningfully closer  to realizing the future of storytelling."  As they proudly stated in many  places, their research team here is led by one   of the co-authors that published a Stable  Diffusion paper. And now that we have decent   models for language and for images, obviously the  next step is video. And honestly, this demo is   the best preview I've seen of this yet. And as we  talked about before on the channel, Facebook and   Google have their own text-to-video generators.  By judging purely from the demos here, this is the   first one that I saw that I would actually love  to use on a daily basis. I mean, heck, imagine me   turning into some fire demon like this. Okay, I am not  sure that would add to the video, but eventually   I would find a way to make this work and improve  even this educational content. Okay, now it's time

to talk about the most mind-blowing application  here that I ran into just a few days ago, and this   was actually the reason I had to make this video.  Listen up, it's starting to happen - text to 3D.   Okay, look, there have been some experiments here,  but with this new research paper that has been   published, they showed off the first demo that  made me actually feel like I live in the future.   Just have a look at this. So, you go ahead and you  draw a car at the level of a two-year-old, and you   get a 3D model of a car. Are you kidding me? And  this is only a 2X speed. This is happening almost   live. And let's take a second to appreciate that  this is day zero here. Everything we're looking at   here today is just the very first preview of what  this can do. I don't know if you're old enough,   but if you remember iPhone 1, it was  revolutionary. But even if you're generous,   that thing was maybe five percent of what the  product is today. And if this 3D aware conditional   image synthesis thing is gonna follow a similar  path - which trust me, it will - the video game   industry has steamrolled movies and music over  the last few years. And the games you could create   with this would be on a whole different level.  So, yes, this will progress. We will be getting   products and experiences here over the next few  years, that we can't even dream of right now. Like   just for a second, imagine talking to somebody on  Microsoft Teams about a red car, and the language   model picks up on that and generates a prompt  that it feeds into image generator. That thing now   generates the image of a red car, and technology  like this will be able to generate a video or a 3D   model of that red car almost instantly. Now come  on, if you add AR and VR to the mix, you could   instantly place that in the meeting room, and  just by talking about an idea, it would instantly   materialize in front of you. So yeah, this  whole AI thing is going to change even more things   in industries than you think right now. Only thing  we can do is stay informed, educate ourselves,   and use this text to our advantage. And that's  what this whole channel is about. I have a   separate video here talking about seven non-hyped  AI trends that are coming up over the course of   the next year, which if you're still here, I  have no doubt you would enjoy. See you there.