UniFab

Description

DVDFab Software launched UniFab as a standalone AI video processing tool, decoupling the neural network-based video enhancement tools from DVDFab’s existing disc ripping and format conversion product line. Where traditional video converters use fixed encoding parameters to generate output, UniFab feeds source video through trained deep learning models that analyze content frame by frame — upscaling low resolution footage towards higher output resolutions, reducing noise from dark scenes and old recordings, interpolating more frames to increase frame rates, and converting standard dynamic range video to HDR output.

The application is aimed at users who would like to enhance existing video quality without the technical complexity of setting up the encoding parameters manually, and instead use AI processing to perform the analysis and enhancement for them, while the user only has to select the output target and quality level.

KEY FEATURES

AI Video Upscaling

The upscaling module is used to increase the resolution of videos by neural network processing instead of using traditional bicubic or lanczos interpolation. DVDFab trained the underlying model on large datasets of video pairs, teaching it to infer fine detail from low-resolution content rather than just making existing pixels larger. Upscaling paths are SD to HD (480p to 720p or 1080p), HD to 4K (1080p to 2160p), and 4K to 8K. The output enhances edges, recovers texture detail, and reduces the blurriness that standard upscaling produces. Processing speed depends on whether a compatible Nvidia, AMD or Intel GPU is handling the acceleration – GPU-accelerated upscaling processes significantly faster than CPU-only processing.

AI Noise Reduction

The noise reduction model eliminates video noise — the grain and speckle that appear in video shot in dim light or compressed heavily during the encoding process — without softening edges and texture that aggressive noise reduction tends to do. The model separates noise from real image information, focusing on the random frame-to-frame variation of noise and leaving consistent image content intact. Old digitized recordings, surveillance footage and heavily compressed streaming rips benefit from this processing.

AI Frame Interpolation

Frame interpolation creates new frames between existing frames in order to increase the output frame rate. A 24fps film is converted to 60fps by generating synthetic frames to represent intermediate states of motion between each pair of original frames. The AI model uses the motion vectors between frames to create realistic intermediate positions for moving objects, resulting in a smoother motion than the frame duplication used by simple frame rate conversion. Sports footage, action scenes and anything with fast motion really benefits from motion compensated interpolation.

AI HDR Conversion

The HDR conversion module takes standard dynamic range video — mastered for 100 nits peak brightness on a standard display — and maps it to HDR output formats including HDR10 and Dolby Vision. The conversion takes the tonal range of the original content and expands it to take advantage of the increased brightness and shadow detail that HDR displays are capable of showing, instead of simply increasing exposure across the board.

Video Conversion and Format Support

Beyond AI enhancement, UniFab supports video format conversion by a traditional encoding pipeline that supports input video formats such as MKV, MP4, AVI, MOV, FLV, etc. Output video encoding formats include H.264, H.265/HEVC, AV1, and ProRes. Conversion is independent of AI enhancement or in combination, applying enhancement processing and re-encoding in a single pass.

Audio Enhancement

An AI audio processing module helps to reduce noise from the recording and improve the clarity of the dialog. The audio enhancement works on the audio track extracted from the video and applies noise suppression trained to isolate the speech from the ambient noise, then remixes the processed audio with the video output.

Batch Processing

Multiple video files are queued for sequential processing with shared or individual settings per file. Each file in the batch displays its processing status, estimated completion time and output file path. Batch processing enables overnight processing of large libraries without the need for manual intervention between files.