Thursday, November 10th, 2022 Posted by Jim Thacker

Topaz Labs ships Video AI 3.1

Originally posted on 10 November 2023. Scroll down for news of the Video AI 3.1 update.

Topaz Labs has released Video AI 3, the latest version of its AI-based software for upscaling, retiming and stabilising video, previously known as Video Enhance AI.

The 3.0 release rewrites Video AI from the ground up, making it possible to apply multiple AI models simultaneously, and adding a new super-slo-mo AI model and support for 16-bit colour depth.

(Full disclosure: the software was actually released last month, but we haven’t seen much coverage of it outside of general video production sites.)

An AI-based tool for upresing video footage with potential applications to animation workflows
Originally released in 2020, and previously known as Video Enhance AI, Video AI is an AI-trained tool capable of a range of video-processing tasks, including upresing, retiming and stabilising footage.

While its main role is in video production, particularly to make poor-quality footage suitable for broadcast, or to upres old material for modern higher-resolution displays, it also has potential uses in CG work.

Many visualisation artists use its still-image counterpart, Gigapixel AI, to upscale low-res 3D renders, making it possible to generate hi-res images more quickly than simply rendering at full resolution.

Unlike Gigapixel, Video AI doesn’t have a dedicated mode for CG imagery, but it does upscale video footage to up to 8K, opening up the possibility of similar workflows for CG animation.

New in version 3.0: stack AI models, run tasks in parallel, and export at 16-bit colour depth
As well as giving it a new name, Video AI 3.0 rewrites the software “from the ground up” to improve its handling of variable frame rate footage, frame repeats, and scene changes.

New capabilities include the option to stack multiple AI models – for example, to upscale and stabilise footage simultaneously – and a new ‘Apollo’ AI retiming model for generating super-slo-mo footage.

Workflow improvements include the option to run tasks in parallel, making it possible to export multiple projects simultaneously; and a command-line interface, primarily for executing ffmpeg commmands.

It also introduces support for 16-bit colour depth and improves “colorspace support and color accuracy”.

Updated 19 January 2023: Topaz Labs has released Video AI 3.1.

The update adds a new AI model for reducing motion blur in footage caused by panning, rotating or zooming the camera, and improves performance by “50-100%” when previewing or exporting processed footage.

Pricing and system requirements
Video AI 3.1 is compatible with 64-bit Windows 10+ and macOS 10.15+. It is compatible with AMD, Intel Arc and Nvidia GPUs. The macOS version supports Apple Silicon natively. It costs $299.

Read an overview of the new features in Video AI 3.0 on Topaz Labs’ blog

Read a full list of new features in the online changelog