Topaz Labs will add generative AI video technology to its Video AI software, allowing users to upscale low-resolution videos to higher resolutions while adding realistic details expected of the objects in the scene.
The generative AI model is trained on many images of the world. Because of this, it knows what object details are present in reality. For example, it knows that a cat has fine whiskers near the nose. Such fine details are not present in low-resolution videos, and traditional upsampling algorithms cannot recreate these lost details.
Project Starlight leverages an AI diffusion model for video to recreate expected, missing details during upscaling, resulting in high-resolution videos that are rich with detail. The feature also operates without artifacts such as incorrect object motions, a temporal error that can occur when AI doesn't create realistic output.
The AI feature will be available next week in the company's Video AI software as a free research preview. Output will be initially limited to processing three ten-second clips per week to 2K resolution for free, while clips up to five minutes will require payment. The full feature set, which will include 4K output resolution, will be released at some point thereafter.
The current Video AI version 6 costs $299, but pricing with the new Project Starlight will be released later.
Source(s)
Topaz Labs unveils Project Starlight, the first diffusion AI model for video enhancement
News provided by Topaz Labs
Feb 13, 2025, 18:19 ET
Restoring previously "impossible" footage to 4K with unprecedented visual quality
DALLAS, Feb. 13, 2025 /PRNewswire/ -- Today, Topaz Labs formally announced Project Starlight, a groundbreaking tool that transforms old, low-res, or degraded video into HD quality. It's the first-ever diffusion AI model for video enhancement and the only one to achieve full temporal consistency, ensuring seamless motion from frame to frame.
Topaz Labs' Project Starlight restored footage example of NASA rocket.
Topaz Labs' Project Starlight restored footage example of NASA astronaut placing helmet.
Topaz Labs' Project Starlight restored footage example of NASA footage of woman watching rocket launch.
Thousands signed up for early access - 1M+ views on X.com - Topaz Labs' Project Starlight
Project Starlight uses diffusion AI to upscale, enhance, denoise, de-alias, and sharpen video—without any sliders or manual adjustments. While OpenAI's Sora and Runway use diffusion technology to generate new footage, Topaz Labs has optimized it—for the first time ever—for video restoration.
"We've been working on image and video enhancement for over 20 years now, and in all that time, there was one main problem with video enhancement—that of temporal consistency," says Topaz Labs' CEO Eric Yang. "Even though we can make one frame look really good, as soon as you put those frames together, you can easily see motion artifacts and inconsistencies between frames and objects. But now, for the first time ever, we're introducing Project Starlight, the world's first diffusion model for video enhancement that fully solves temporal consistency. With it, you can get surprising results on everything from old home videos to professional shoots. We consider this one of the most important breakthroughs we've ever made."
Early tests confirm Starlight's capabilities. When the company shared initial results online, the posts reached 1M+ views (and growing) with thousands of users requesting early access.
Launch Details
Project Starlight is available to consumers, professionals, and enterprise clients. Anyone able to upload a video can preview its capabilities.
A free research preview launches next week in the Video AI desktop app, allowing users to process three 10-second clips per week via the cloud. Upscaling will be limited to 1080p for the free preview with 4K access launching in the future. If satisfied with the preview results, users will be able to purchase extended cuts up to 5 minutes in length. Pricing will be announced at that time.
The full, paid version is set to launch later this year, with local processing a goal for the future.
"Starlight is just the beginning," says Video AI Product Manager, Anthony Martinez Cuervo. "It's the first of an entirely new series of models, and we see a clear path toward improving Starlight's quality and speed even beyond this initial release."
To learn more…
Try the Project Starlight research preview at app.topazlabs.com/experiments/starlight
Enterprise inquiries: [email protected]
About Topaz Labs
Founded in 2005, Topaz Labs is a leader in AI-powered image and video enhancement. Its technology is used by 1.5 million customers, including 20 of the world's top 50 companies. For more information, visit www.topazlabs.com.
SOURCE Topaz Labs