Dr. Zsolnai reviews an AI from Wiseman Institute and Nvidia that edits video and images with text prompts. You can change backgrounds, swap clothing while tracking wrinkles, transform objects like cake into ice cream, or add effects like fire to specific regions. The results sometimes lose fine detail.
Dear fellow scholars, this is Two Minute Papers with Dr. Zsolnai. Today we are going to have a look at this magical AI from scientists at the Wiseman Institute of Science and Nvidia, which promises magical video editing powers. So let's see four of my favorite things it can do together. One, we can make this unremarkable video of a car passing by. And my goodness, look at that. It has completely revamped the scene. First, it made sure that it is nighttime and has given the car a nice neon cyberpunk look. So this took quite a bit of work for someone, right? Well, not quite. To achieve this, all we need to do is give the AI a video and two text prompts. One for the foreground and one for the background. And if we are looking for an example that is a little less out there, we can also ask that the car is made rusty instead. So cool. Two, Once again, if we did not show up in the right outfit for a day, or would like to see what we would look like wearing something else before buying it, look no more. Yes, this is also just one text prompt away. I particularly like this example because this outfit has lots of wrinkles and geometry changes over time, and the new technique can follow them really well. Now, if you look closely, you see that there is also a price to be paid for this look. The new synthesized results are not as detailed as the original video. Now, we said that it can do foreground and background editing separately, and that was the foreground. But what about the background? Oh yes. Let's make this environment a little foggier. Look at that. I love this one. Even the shadows have become foggier. So cool. And the shadows are even taken away for the nighttime shot because of the different lighting setup. I would like these results a great deal for images, but for video. Such an amazing result for video. Wow. Now, of course, this kind of layered editing concept can be applied to images too, and the results are outstanding. For instance, three, if we have this boring cake, we can reimagine it if it were made of Oreos instead. Or if we wish to let our imagination run wild, or we are light transport researchers, we can ask the AI to make it out of ice. Loving it. Now, if you aren't hungry enough yet, here is a piece of bread that can be converted to the legendary red velvet bread. 2. Or once again, ice. And 4, the authors call these semi transparent effects. One, where we can choose a small part of an already existing image, add the text prompt, for instance, fire. And then bam. We suddenly have a fire breathing bear or a beautiful cup of coffee. This technique can truly do so much. Now I would like to send a huge thank you to the authors of the paper because the source code for this project is also available so AI driven layered image editing is possible for free and for everyone. And just imagine what this technique will be able to do just a couple more papers down the line. What a time to be alive. So what do you think? What would you use this for? Let me know in the comments below. This video has been supported by Weights and Biases. Look at this. They have a great community forum that aims to make you the best machine learning engineer you can be. You see, I always get messages from you fellow scholars telling me that you have been inspired by the series but don't really know where to start. And here it is. In this forum you can share your projects, ask for advice, look for collaborators and more. Make sure to visit wnb me paperforum and say hi or just click the link in the video description. Our thanks to Weights and Biases for their long standing support and for helping us make better videos for you. Thanks for watching and for your generous support and I'll see you next time.
