Topaz Video AI Paid
Desktop app that upscales and de-noises video using local GPU models, up to 8K and 60fps
Topaz Video AI is a desktop application that upscales, de-noises, stabilizes, and frame-interpolates video using local GPU models, with output up to 8K. It costs a one-time fee instead of a subscription, and it will test your graphics card.
What is Topaz Video AI?
Made by Topaz Labs in Dallas, the app replaces the older Video Enhance AI line. You drop a clip in, pick a model (Proteus, Iris, Rhea, Nyx, and others), set an output resolution, and render. Everything runs locally, which means no cloud upload and no per-minute billing, but a 4K upscale of a 10-minute clip can take over an hour on a mid-range card.
Current version is 6.x. The models are trained per category: Iris targets faces, Nyx handles noise, Rhea is the newer general-purpose upscaler with better speed.
Product Features
- Model library with dedicated networks for faces, noise, compression artifacts, and slow motion; you can stack two models in one pass.
- Frame interpolation up to 120fps with the Apollo model, useful for sports and archival footage.
- Local GPU processing with NVIDIA, AMD, and Apple Silicon support; VRAM usage scales with resolution and output size.
- Batch queue that renders a folder of clips overnight with consistent settings.
- Export presets for H.264, H.265, ProRes, and image sequences.
Product Highlights
Real results on old footage. DVD-era 480p material came out visibly cleaner, particularly faces. On already-sharp 1080p source, the gain was marginal and sometimes added a plastic look.
Speed depends entirely on hardware. A 10-second 720p to 4K test rendered in 40 seconds on an RTX 4070 and about 6 minutes on an M1 MacBook Air. Laptops are not the target machine.
The upgrade path is the sore point. The perpetual license includes 12 months of updates. After that, new models cost extra. Users who bought in 2023 had to pay again for the Rhea models in 2025.
Use Cases
- Restoring family archives and old broadcast tapes before they degrade further.
- Upscaling client footage shot on older phones to meet a 4K delivery spec.
- De-noising high-ISO drone or night footage before color grading.
- Creating slow motion from 24fps source without visible stutter.
- Preparing legacy training videos for modern LMS platforms.
How much does Topaz Video AI cost?
The perpetual license is $299, with a $99 per year upgrade option to keep new model drops. There is no subscription-only plan, and no free tier beyond a watermarked trial. For a studio processing hundreds of clips a month, that math beats most cloud upscalers quickly.
Don’t buy it for phone-shot social clips; a cloud API costs cents and you avoid the render wait. Do buy it if you handle sensitive footage or process volume regularly.
Quick Summary
- Topaz Video AI is a local desktop app for upscaling, de-noising, and interpolating video up to 8K.
- It costs $299 for a perpetual license, with $99 per year to keep receiving new models.
- Rendering is GPU-bound: 40 seconds on an RTX 4070 versus about 6 minutes on an M1 MacBook Air for the same clip.
- Best for archivists, post houses, and anyone with sensitive or high-volume footage.
- Overkill and slow for casual short-form video work.
