Enhance video quality with the new Video Enhance AI Model Manager
Published On: August 11th, 2021
Video Enhance AI is the best solution to enhance videos in a variety of ways, including upscaling a low-resolution video up to 8K, performing a frame rate conversion on your video clips, or applying up to a 2000% slow-motion effect. We’re also continually looking for new ways to help you enhance video quality using our growing collection of AI models.
As we’ve grown both the number of AI models and the versions of each model in Video Enhance AI, we’ve listened closely to the helpful feedback our users have provided. One item that stood out involves offering better management of these AI models. Some of our AI models have pretty large file sizes that can have a real-world impact on users, especially those who have limited internet bandwidth to download them or do not have enough hard drive space to locally store them. That’s why we’re excited to announce the AI Model Manager in Video Enhance AI v2.4.
The right AI model at the right time
Do you have a one-off low-resolution video that suffers from aliasing? The AI Model Manager allows you to control which models you want to download on the fly. You can use it to download the Artemis Aliased & Moire model, process your video using it, and remove it until the next time you need it.
We’re very excited about the opportunities that the AI Model Manager affords our users. We recently spoke with Taylor Bishop, our lead developer, to discuss why we built it and its meaning for our future app and AI model development. But, before we dive into that interview, we’d like you to check out this brief video tutorial illustrating how to use the AI Model Manager in Video Enhance AI.
Why we built the AI Model Manager
The new AI Model Manager has far-reaching implications in how we approach both application and model updates. To explain a bit more about it, we’d like to share this interview with Taylor.
How about we start with some formalities? Can you tell us who you are and what you do at Topaz Labs?
Hi, everyone! My name is Taylor Bishop. I’m the lead developer on both Video Enhance AI and Gigapixel AI. My core focus is ensuring that we present users with the AI technology that underpins both applications in the easiest to use way possible while allowing for as much flexibility as makes sense for our end users. My ultimate goal is to have every user, no matter what type of image or video they’re looking to enhance, be able to look at the result from our products and say “wow.”
Thanks for the introduction, Taylor. The primary theme of this article is to introduce the new AI Model Manager within Video Enhance AI. What led to the creation of this new feature?
We’ve been exploring ways to serve our users best regarding AI model management within Video Enhance AI. We built the AI Model Manager to strike a balance between ease-of-use and flexibility for our users. We kept asking ourselves, “How do we deliver new models to users without overloading them with options? How do we let people experiment but not overload those that want 1-click functionality?”
The goal was to allow users to dig in, experiment, and decide where to store our growing library of AI models. We also want to provide the user with control over how much of each model will be stored locally without interrupting the core user experience.
How often do you expect to release new model classes and versions?
We’ve been growing our R&D team in recent months and hope to be putting out new model versions at least once every 1-2 months. Our R&D team is constantly experimenting with bleeding-edge techniques in the machine learning field, and our giant pseudo-supercomputers are chugging away every day to train those models. Even if we haven’t actively changed the guts of one of our models, our computers are constantly training them, allowing us to deliver that value quicker than before.
I recommend reading this thorough article to learn more about how we train our AI models to improve the objective quality of still images and videos. The article was written by Partha Acharjee, one of our brilliant research engineers. He explains how we use AI to apply intelligent noise reduction to still images.
What does the AI Model Manager mean for the rest of the Topaz apps?
The AI Model Manager—and more importantly, the technology that underpins it—allows us greater flexibility in delivering new models and model updates without bogging users down with massive amounts of data. It provides flexibility and choice for users to decide if they want to download every model we have available and then go offline or have the product pick and choose which parts of a model to use for any given video and scale. It’s a way of providing our more ambitious users a choice while not bogging down the average user with options.
We’ve essentially replaced the plumbing and are actively disseminating this method of model management to our other products and put a window into this with the Model Manager.
Thanks so much for sharing these insights with us, Taylor! Before we wrap things up, I have one more critical question. What’s your favorite flavor of ice cream?
That question plagues every person at least once in their lives and has personally kept me up at night on multiple occasions. Cookies & Cream or Chocolate Chip Cookie Dough? Ben & Jerry’s Cookie Dough Core or Heath Bar Crunch? The older I’ve gotten, the more comfortable I’ve become with the notion that I may never have a definitive answer to this question. What is life but a series of questions we ask ourselves, often with very few solutions?
That’s true, Taylor. Although, the only correct answer is Mint Chocolate Chip.
If you’re as excited about the new AI Model Manager and want to try it out for yourself, download the free trial today! Also, don’t forget to leave a comment below to share your thoughts on Video Enhance AI and the new AI Model Manager.
Brian Matiash serves as Product Marketing Manager for Topaz Labs and manages the Topaz Labs Learning Center. He is also a photo educator and author, with his work being featured in dozens of international publications. Learn more about Brian by visiting his website.