All mirrors online | Library checked 6 hours ago | 249 titles indexed | 131.5M total downloads served
T trpger.us
Home / All software / Video & Motion / Topaz Video AI Pro
Topaz Labs, Video & Motion

Topaz Video AI Pro

Pre-ActivatedModels IncludedOffline InstallerVersion 7.1.664-bit

Model driven video restoration that upscales, denoises, deinterlaces and interpolates frames, aimed at old or compressed footage that conventional filters cannot recover.

Version
7.1.6
Package size
2.87 GB
Downloads
869,484
Licensing
Pre-activated
Architecture
64-bit
Updated
5 days ago
Rating
4.4 / 5
Votes
7,466

About this release

Topaz Video AI is a restoration tool rather than an editor. You feed it footage that is too small, too noisy, interlaced, or shot at the wrong frame rate, and it runs a trained model over every frame to produce something usable. On archive material, old camera footage and heavily compressed sources, the result is well beyond what a sharpening filter can do.

Upscaling is the headline. The models reconstruct detail rather than interpolating pixels, so standard definition source can be taken to high definition or beyond with detail that reads as real rather than smeared. Different models suit different sources, so there is a compare view that renders the same section through several of them side by side before you commit to a full pass.

The rest of the toolset handles the other common problems. Denoise for grain and compression artefacts, deinterlace for broadcast and tape sources, stabilisation with full frame reconstruction that fills the edges instead of cropping in, and frame interpolation for smooth slow motion or for converting between frame rates. Passes can be chained, and batch processing runs a queue across a whole folder.

What it does

  • Model based upscaling that reconstructs detail rather than interpolating
  • Multiple models suited to different source types with a side by side compare view
  • Denoise for film grain, sensor noise and compression artefacts
  • Deinterlacing for broadcast and tape sourced footage
  • Stabilisation with full frame reconstruction rather than cropping
  • Frame interpolation for slow motion and frame rate conversion
  • Chained passes so several operations run in a single render
  • Batch processing across a whole folder with a queue
  • Preview render of a selected section before committing to a full pass
  • Export to editing codecs including ProRes and image sequences

Changes in this version

  • Updated model set with better handling of heavily compressed sources
  • Faster processing on current generation GPUs
  • Improved stabilisation edge reconstruction
  • Better memory handling on long clips at high output resolution
  • Reworked compare view with synchronised scrubbing

What the package contains

  • Topaz Video AI Pro offline installer
  • Complete model set, no download required on first run
  • Export presets for editing codecs and image sequences
  • Sample project for testing model choices
  • Readme with the install order written out step by step

Installation notes

Follow these in order. Most reports of a failed install come from skipping a step rather than from the package itself.

  1. Disconnect from the network.
  2. Extract to a short path with plenty of free space.
  3. Run the installer as administrator.
  4. Place the included model set where the readme indicates before first launch.
  5. Launch once and confirm the model list is populated.
  6. Keep the network off, the application does not need it once the models are in place.

System requirements

Operating systemWindows 10 version 22H2 or Windows 11, 64-bit
ProcessorIntel or AMD 64-bit with AVX2 support
Memory16 GB minimum, 32 GB recommended
GraphicsGPU with 6 GB VRAM minimum, 8 GB or more strongly recommended
Storage20 GB free plus space for output, which is large
Display1920 by 1080 or larger

Things worth knowing

  • Processing is slow by nature. A few minutes of footage can take hours depending on the model and output resolution.
  • Output files are very large, especially to ProRes. Check your free space before starting a batch.
  • Test model choices on a short section with the compare view before running a full pass, they differ a lot by source.

Questions about this title

How long does a pass take?
Depends entirely on model, source resolution and output resolution. Budget hours rather than minutes for anything beyond a short clip.
Which model should I use?
There is no universal answer, which is why the compare view exists. Old tape, film scans and compressed web video all favour different models.
Does it need a network connection?
No. The complete model set is included, so nothing streams on first use.
Can it fix any footage?
It works with what is there. Severely damaged or extremely low resolution source has limits, and pushing past them produces artefacts that look worse than the original.

What people are saying

Listing information

This entry is catalogued under Video & Motion and carries the tags below. Language coverage is English, and the supported platform list is Windows 10, Windows 11. Version 7.1.6 is the current catalogued build, last checked 5 days ago, and the entry has been in the index since it was added 2 years ago.