Topaz Video AI

Footage that is degraded or low-resolution goes through a diffusion-based upscaling process that produces clean, consistent HD and 4K output. Rather than treating each frame in isolation, the model processes neighboring frames together. This keeps detail stable across motion and complex content without any manual tuning.

Eight specialized models sit inside Topaz Video AI, each built for a different footage type or enhancement challenge. Noise removal, face restoration, sharpening, and texture upscaling are all covered at varying intensity levels. The right model depends on the source condition and quality of the footage being processed.

New frames are generated between existing ones using a set of purpose-built AI models for frame rate conversion and slow motion. Apollo handles complex or blurry motion, Chronos is suited to linear movement, and Aion is designed for high-resolution input with large movement. Each model targets a specific type of motion challenge.

Standard dynamic range video gets expanded contrast, brightness, and color depth to produce HDR output. The conversion works entirely from the color information already present in the source clip. No HDR footage is needed as a starting point.

Shakiness and jitter in recorded footage are reduced by analyzing camera movement across frames. The process works on footage that has already been shot, with no reshooting required. It can also be applied independently from any of the other enhancement models.

An adjustable film grain layer that sits on top of existing footage. The texture is built to look like natural film grain. It can be combined with other AI enhancement models to produce a result that looks less digitally processed.

Blur caused by panning, zooming, or fast-moving subjects is addressed through a dedicated sharpness restoration model. Unlike general sharpening, the focus here is specifically on movement-related blur. This makes it distinct from the standard enhancement models in the library.