Project Starlight by Topaz Labs revolutionizes video enhancement with AI-powered diffusion technology, delivering HD quality and smooth motion.
Topaz Labs launches Project Starlight, a groundbreaking AI tool using diffusion technology to upscale and restore videos with full temporal consistency. Image: Topaz LabsDallas, Texas, USA — February 14, 2025:
On Thursday, Topaz Labs introduced Project Starlight, an innovative tool that brings old, low-resolution, or degraded video to life in stunning HD quality. This groundbreaking technology marks the first-ever diffusion AI model designed specifically for video enhancement, and it promises to solve a long-standing challenge in the industry: achieving full temporal consistency. This ensures that motion remains smooth and artifacts are eliminated between frames.
Project Starlight leverages diffusion AI to upscale, enhance, denoise, de-alias, and sharpen video automatically, without requiring manual adjustments. Unlike other technologies like OpenAI's Sora and Runway, which generate new footage using diffusion models, Topaz Labs has fine-tuned this technology for video restoration.
"We've spent over 20 years working on image and video enhancement, and solving temporal consistency in video has always been the major hurdle," said Eric Yang, CEO of Topaz Labs. "Project Starlight is a breakthrough, solving this issue and delivering exceptional results on everything from old home videos to professional footage."
The tool is now available for preview, allowing both consumers and professionals to upload videos and test its capabilities. A free research preview will be available next week via the Video AI desktop app, where users can process up to three 10-second clips per week in the cloud. For the free preview, upscaling is limited to 1080p, with 4K access expected to roll out later. If users are satisfied with the preview, they can purchase extended clips up to five minutes in length, with pricing to be announced.
The full paid version of Project Starlight will launch later this year, with plans for local processing in the future.
"Starlight is just the beginning," said Anthony Martinez Cuervo, Video AI Product Manager. "We see tremendous potential in refining this technology further, improving both quality and speed in future releases."