Upscaling, denoising, deinterlacing, stabilisation and frame interpolation driven by trained models, with every model included offline.
Topaz Video AI does one category of thing and does it better than anything else available: taking video that is not good enough and making it usable. Old footage, low resolution captures, interlaced broadcast material, compressed downloads, handheld shots and clips that need to run at a different frame rate all go in one end and come out substantially improved.
Upscaling is the headline. Rather than interpolating pixels, it uses models trained on the relationship between low and high resolution footage to reconstruct detail that a conventional scaler cannot produce. Standard definition to high definition works genuinely well. High definition to 4K is convincing on most material. There are separate models tuned for compressed sources, for clean digital sources, for interlaced broadcast, and for computer generated content, because each degrades in a different way.
The other tools are as useful in practice. Frame interpolation generates intermediate frames for smooth slow motion or frame rate conversion, without the tearing that optical flow produces on difficult motion. Stabilisation works on full frame or rolling shutter footage with adjustable cropping. Denoising removes sensor grain and compression blocking without smearing the detail underneath. Deinterlacing produces progressive frames from broadcast material properly.
It works standalone or as a plugin, processes in batch, previews any model on a section of the clip before committing, and this build carries every model file offline so nothing has to be downloaded on first use.
Read the whole list before you start. Most of the reports that come into the desk about this title turn out to be a step taken out of order.
| Processor | Intel or AMD, 64 bit, six cores or more recommended |
|---|---|
| Memory | 16 GB minimum, 32 GB for 4K work |
| Graphics | 8 GB VRAM strongly recommended, processing is almost entirely GPU bound |
| Storage | 10 GB for the install and models plus space for output |
| Display | 1920 by 1080 |
Processing is slow by nature. A few minutes of footage can take hours depending on the model and the target resolution.
Preview the model on a short section first. Choosing the wrong model wastes an entire render.
| Title | Topaz Video AI Pro |
|---|---|
| Developer | Topaz Labs |
| Version | 7.1.6 |
| Category | Video & Audio |
| Release type | Pro (full) |
| Operating system | Windows 11, Windows 10 (64 bit) |
| Architecture | x64 |
| Interface language | English |
| Archive size | 2.86 GB |
| File name | topaz_video_ai_pro_7.1.6_x64_full.rar |
| Shelf updated | 5 days ago |
| Total downloads | 869,484 |
The timeline that most of the industry cuts on, with the text based editing pass, native multiformat handling and the whole audio workflow intact.
Motion graphics and compositing with the full 3D workspace, expressions engine, Roto Brush and the plugin folder left open for the third party packs everyone runs.
A fast, direct video editor with auto captions, background removal and a large template and effects library, unlocked to the full feature set.
Editing, colour, visual effects and audio post in one application, with every Studio only feature including the neural engine tools and full resolution output.