Topaz Labs has released its Precision Update, adding four new models and nine improvements across its image and video enhancement suite. For AI video producers, the headline additions are two new upscale models and extended AMD GPU support that closes a gap for a significant portion of the professional user base.
Starlight Precise 2.5 is the more immediately relevant of the two new video models. Built to bring generative AI video to 4K, it is tuned specifically for the artefact patterns that current text-to-video tools produce – the temporal inconsistencies, texture problems, and motion blur that cause AI-generated footage to fail broadcast and streaming delivery specs. As we examined in The Last Mile Problem, broadcast and streaming platforms require 4K UHD progressive as a minimum under IMF mezzanine delivery standards, and most generative tools currently top out at native 1080p. Starlight Precise 2.5 is designed to close that gap in a single finishing step.
The second new video model, Gaia 2, is optimised for animated and graphic content rather than live-action generative footage. Animated material has different artefact characteristics to live-action, and the separation reflects a more considered product approach than offering a single model for all content types.
Also in the release is local support for the Wonder 2 model on AMD GPUs, enabled by Topaz NeuroStream, a proprietary VRAM optimisation technology. Wonder 2 handles upscaling, sharpening, and denoising without requiring parameter adjustments, and the AMD addition extends that capability to hardware that has until now been excluded from local model access. Topaz also describes the update as its largest API release to date, pointing towards deeper integration with automated production pipelines.
The Precision Update is available now. Full details at topazlabs.com.
