Need product and account support?

Before
After
Drag to reveal
New at topaz labs

Local Update

Run our largest, most powerful models locally, even on consumer hardware.

Our team’s new proprietary, industry-first inference system reduces VRAM usage by up to 95%. This allows creators to run Wonder 2 right on their own machines.

Key models

Wonder 2 (Local)

Topaz Photo

Starlight Fast 2

Astra
Other Features

Scene Controls

Astra

Batch Rendering

Astra
Key Models
Before
After
Compare
New Model

Wonder 2 Local

Get the same high-accuracy results with Wonder 2, now available to run locally. Restore every detail with precision upscaling in low-resolution images without oversharpening outputs.

Example 01

Example 02

Example 03

Before
After
Compare
New model

Starlight Fast 2

Achieve superior results 2x faster. This next-gen model—successor to Starlight Speed and Sharp—is optimized for processing speed, and enhances everything from medium and high-quality generative content to archival footage.

Example 01

Example 02

Example 03

Example 04

Introducing Topaz NeuroStream

Run the largest, most powerful AI models. Locally.

Topaz NeuroStream dramatically reduces VRAM requirements for large AI models, cutting usage by up to 95%. This makes professional-grade image and video editing accessible to everyday prosumers, and lays the groundwork for running other powerful AI models locally on consumer-grade GPUs in the future.

Large Video Model

56GB
VRAM Used

Topaz NeuroStream

2.8GB
VRAM Used
Other Features

Scene Controls

Astra

Upload a single video file and apply different enhancement modes to individual scenes to streamline your workflow in Astra.

Batch Rendering

Astra

Save time by batch uploading and processing footage that benefit from the same enhancements and settings.