Topaz Video

Topaz Video AI is a desktop application that employs neural networks to increase video resolution, remove compression artifacts, recover details from low-quality footage, and generate intermediate frames. The tool operates with models specifically trained for video, distinguishing real noise from original texture. It processes clips up to 16K, stabilizes shaky recordings, converts frame rates between 24, 60, and 120 fps, and creates slow-motion effects without visual loss. The software is optimized for NVIDIA, AMD, and Intel hardware through GPU acceleration.

★★★★★

1.0(1 ratings)

• Resolution upscaling with Gaia-HQ and Artemis. Gaia-HQ is a model focused on preserving natural edges and original textures while scaling video up to 4K, 8K, or 16K. Artemis specializes in reconstructing fine details in compressed or low-resolution content, restoring sharpness without introducing artifacts. Both models analyze each frame independently and apply millions of parameters learned from pairs of low-quality and high-quality videos.
• Chronos model for frame rate conversion. Chronos is a temporal inference model that converts videos from 24 fps to 60 fps and can elevate them to 120 fps or beyond. It generates intermediate frames by synthesizing motion between existing frames, producing smooth transitions in scenes with fast movement. It is used for converting cinematic material to television standards or creating slow-motion sequences without basic interpolation.
• Aion model for high-quality slow motion. Aion is specifically designed to generate additional frames in shots with fast or complex motion, such as sports or action sequences. Unlike Chronos, Aion focuses on maintaining coherence of foreground and background objects, reducing ghosting blur. It allows slowing videos to 1/8 of their original speed while preserving fluidity and detail.
• Compression artifact and block removal. This function detects macroblocks, halos, and color banding resulting from high-compression codecs (such as H.264 at low bitrates). It uses a neural network trained to replace damaged areas with textures similar to the original video, recovering smooth gradients and reducing the staircase effect on diagonal edges.
• Face restoration and facial recognition. The software includes a dedicated model to identify faces in video, even in distant shots or with visible pixels. It restores eyes, nose, and facial contours, corrects specific blur on skin and hair areas, and unifies facial lighting. This function applies selectively without affecting the rest of the scene.
• Video stabilization through motion analysis. Analyzes pixel trajectories between frames to separate intended camera movement from unwanted shake. It calculates a smoothed motion curve and repositions each frame to compensate for vibrations, jitter, or sudden movements. Useful for recordings with unstable handhold, drone footage, or tripod shots with defects.
• Temporal and spatial noise reduction. Combines two strategies: spatial reduction analyzes pixels within the same frame to remove grain and color noise; temporal reduction compares neighboring frames to separate random noise from actual detail. Levels can be adjusted separately and the system preserves fine textures such as skin, grass, or fabric.
• Motion blur correction. Detects motion vectors within the scene and applies adaptive deconvolution to reverse blur caused by moving subjects or fast camera movement. Recovers sharp edges on moving objects without artificially increasing noise in static areas.
• Color and dynamic range reconstruction. Expands the color range of videos with low bit depth (8-bit) by simulating a 10-bit or 12-bit range. Recovers smooth color transitions in skies, shadows, and reflections, corrects color banding, and improves saturation without oversaturating. Compatible with Rec.709 and Rec.2020 color spaces.
• Batch processing and export profiles. Allows loading multiple video files simultaneously and applying the same combination of models (upscaling, interpolation, noise reduction) to all of them. Custom export profiles can be saved with container format (MP4, MOV, TIFF), codec (H.264, ProRes, HEVC), and bit depth. The processing queue executes tasks sequentially using allocated GPU resources.
• Proteus model for manual fine-tuning. Proteus exposes parametric controls such as sharpness enhancement, noise reduction, fine detail recovery, and halo suppression. Unlike automatic models, it allows the user to modify four dials in real time while observing a preview. Useful when the source video has varying defects across different scenes.
• Frame interpolation for doubling frame rate. Generates new intermediate frames using optical flow estimated by a neural network. Can either double the exact frame rate (e.g., 30 to 60 fps) or set any fractional ratio (e.g., 24 to 50 fps). Maintains original audio synchronization and proportionally adjusts the audio track length.
• Live preview with split comparison. Before exporting, the software shows a viewer divided into two parts: left side original and right side with activated models. A vertical line can be moved to compare any area of the frame. The preview is calculated in seconds due to GPU optimization, allowing rapid iteration of Gaia, Chronos, or Proteus combinations.

Topaz Video AI began development in 2018 under the name Topaz Video Enhance AI, created by Topaz Labs, a company founded by Albert Yang and his team based in Dallas, Texas. The first stable public version was released in August 2019. The software is written predominantly in C++ for the video processing core and neural network engine, while the user interface and operating system integration uses C# on Windows and Objective-C/Swift on macOS. The artificial intelligence models are trained with Python and frameworks such as PyTorch, exported to a proprietary inference engine. The current version unified all models under the name Topaz Video AI (dropping the Enhance suffix) in 2022, adding Chronos, Aion, and Proteus.

Alternatives to Topaz Video:

UniFab

UniFab Video Enhancer AI is a video processing application that uses artificial intelligence algorithms to improve video quality, increase resolution up to 16K, remove noise, and convert formats.

Price: Free   Size: 8.9 MB   Version: 4.0.2.0   OS: Windows

Topaz Photo

Topaz Photo AI is an application specialized in enhancing the quality of photographic images through artificial intelligence algorithms.

Price: $169   Size: 959 MB   Version: 1.3.2   OS: Windows, MacOS