# Google’s New AI: The Age of AI-Made Movies Is Here! ## Метаданные - **Канал:** Two Minute Papers - **YouTube:** https://www.youtube.com/watch?v=mFzgTcTpqtI - **Дата:** 12.02.2023 - **Длительность:** 6:54 - **Просмотры:** 157,410 - **Источник:** https://ekstraktznaniy.ru/video/13289 ## Описание ❤️ Check out Weights & Biases and sign up for a free demo here: https://wandb.com/papers 📝 The paper "Phenaki - Realistic video generation from open-domain textual descriptions" is available here: https://phenaki.research.google/ My latest paper on simulations that look almost like reality is available for free here: https://rdcu.be/cWPfD Or this is the orig. Nature Physics link with clickable citations: https://www.nature.com/articles/s41567-022-01788-5 🙏 We would like to thank our generous Patreon supporters who make Two Minute Papers possible: Aleksandr Mashrabov, Alex Balfanz, Alex Haro, Andrew Melnychuk, Benji Rabhan, Bryan Learn, B Shang, Christian Ahlin, Edward Unthank, Eric Martel, Geronimo Moralez, Gordon Child, Jace O'Brien, Jack Lukic, John Le, Jonas, Jonathan, Kenneth Davis, Klaus Busse, Kyle Davis, Lorin Atzberger, Lukas Biewald, Matthew Allen Fisher, Matthew Valle, Michael Albrecht, Michael Tedder, Nevin Spoljaric, Nikhil Velpanur, Owen Campbell-Moore, Owen Skarpnes ## Транскрипт ### [] Dear Fellow Scholars, this is Two Minute  Papers with Dr. Károly Zsolnai-Fehér. I have to pinch myself that I am saying  but, but here goes. Today we are going to   use Google’s amazing new AI to create not an  image, and not even a short image sequence,   but we are going to make an entire film. An  AI-made film, and all this within seconds. Yes, that’s right. Less than a year ago, in  April 2022, an amazing text to image AI called   DALL-E 2 was published by scientists  at OpenAI. This took a piece of text,   and painted an image for us. It ushered  in a new era of AI-assisted tools. And then, something amazing happened. And not  6 years later, but about only 6 months later,   in October 2022. This follow up  paper could generate not just images,   but short videos of our choosing,  but even if it did that really well,   this was still made to create only that…short  sequences. A true miracle of AI research. However, here is another one. And  hold on to your papers, because now,   this new technique called Phenaki promises no  less than entire films made by an AI. Entire ### New technique [1:24] films! Now I am super excited for this, so let’s  have a look at 5 of my favorite examples together. One. The new work’s superpower is that  we can write several text prompts,   and the AI finds a way to chain them together in  a way that makes sense. That sounds…that sounds   exactly like the script for a movie.   Amazing. If it works, of course. So,   first, we can say “A photorealistic teddy bear  is swimming in the ocean in San Francisco. ” Then, ### Transformation [1:59] we say “The teddy bear goes underwater. ” Then,  “The teddy bear keeps swimming under the water   with colorful fishes. ” And finally,  “A panda bear is swimming underwater. ” And the AI finds out by itself how this  magical transformation from teddy bear   to a panda should take place. I love it. Now,  this was really cool, how about another one? Two. We start out with a side view of an astronaut  who is walking through a puddle on Mars. Who then ### Astronaut [2:35] starts dancing. I am particularly interested in  these transitionary moments, this one checks out,   really cool. Then, he walks his dog on Mars. And  look at that! The dog does not seem to just grow   out of nothing, but it gradually enters the frame.   Once again, a graceful transition. And then,   the astronaut and this good boy watch  fireworks together. Well, kind of. Three, I also loved that we can also give  the AI a little more explicit instructions,   for instance, we can ask it to zoom  out at the end of the video when   the campfire appears. And it did  exactly that. Good job, little AI! ### Zooming out [3:26] Four, the paper also showcases a two and a  half minute long video that starts out with   a scenic motorcycle ride, running through  the woods and finding beautiful things,   of course, robots, and more. And five, while we look through the beautiful  story of this penguin, let’s talk about pixels. ### Pixels [3:51] These are 128x128 videos. That is not a lot  of pixels. Now, of course, these are not super   high-quality videos yet, they do not compare to  the super-high quality images that text to image   AIs are capable of. So why be so excited for this?   Well, we are experienced Fellow Scholars here, ### 5. Penguin story continues [4:18] so what do we do? We apply the First Law of  Papers. The First Law Of Papers says that research   is a process. Do not look at where we are, look  at where we will be two more papers down the line. So, where will we be a couple more papers  down the line? Well, for reference,   this is what DALL-E 1 was capable of,  and just one more paper down the line,   this was possible with DALL-E 2. And looking  at the previous results in image generation,   we are likely having a DALL-E 1 moment for AI  video generation. Yes, this means we are likely   on the cusp of a revolution. Just look at how much  progress can be made in just one paper. Wow. So,   high-quality AI movies are likely not  impossible anymore. And not even decades   away. They might very well be just months  away from now. What a time to be alive! And, if you wish to be part of this amazing  progress in AI research, consider subscribing   and becoming a Fellow Scholar, and even  better, hitting the bell icon to not miss it. Now, plus one, this has become one of my favorite  parts of the paper. Get this, it can even animate   an already existing image of ours given a text  prompt. So, we can not only create entire movies,   but we can even tell the AI what kind of  movie we wish to create. Wow. My goodness. So, what do you think? What would you use  this for? Let me know in the comments below! Thanks for watching and for your generous  support, and I'll see you next time!