Smart Video

Our Video Frame Insertion is a frame rate upconverter utilizing AI deep learning models and optical flow estimation to generate new frames between existing frames. The result is clear and fluid motion that looks more lifelike and immerses the viewer in the video.

Our Super Resolution is a video resolution upscaling technique that can improve subjective image quality while increasing the video resolution. Traditional video resolution upscaling algorithms does not improve the subjective quality as details often look blurred at the higher resolutions. Our Super Resolution uses AI algorithms and machine learning models to analyze each frame and the surrounding frames to increase the resolution while intelligently preserving important details. Our Super Resolution algorithms are optimized to run on GPUs for accelerated processing and support resolution upscaling to 4K, 8K, and more.

Our Smart Enhancement integrates multiple video and image processing modules such as face detection, text detection, texture detail detection, color enhancement, and many more for quality enhancement and the repair of videos or images. Videos using our enhancement algorithm result in more pronounced details for a richer video experience with less processing time.

Night shooting is often problematic due to insufficient lighting reaching the camera sensors, resulting in a video that is often too dark and with muddy details. Our Night Scene Enhancement intelligently improves brightness and object details after the video capture, alleviating the need for more expensive equipment for night shooting.

Lack of clarity in images could result from many issues, including low resolution, hardware limitations, and other reasons. Our Detail Enhancement enhances images with richer details, edge enhancement, and visual clarity.

Our Smart Deblocking uses AI models trained to recognize ugly compression artifacts that affect subjective quality and viewer's enjoyment, such as macroblocking and ringing effects. Our Smart Deblocking algorithm selects the optimal process to repair the video for higher visual quality.

Video jitter (or shaky video) is a quality problem caused by unstable filming scenarios such as handheld filming or vehicle riding. The Debounce capability is a video stabilization technique using software to remove video jitter and improve the viewing experience. Since the process takes place after the original filming, Debounce can be applied to any video regardless of the filming hardware.

Video Denoising uses Media Lab's proprietary algorithms to eliminate noise and improve visual quality. Removing video noise as a pre-processing step greatly benefits video encoding quality and lowers the bitrate requirements for quality targets.

Scratched Snowflakes Removal uses AI machine learning to identify and repair visual quality problems such as film scratches and snowflakes. The resulting video is significantly less distracting to the viewers.

Through intelligent video analysis, Old Film Restoration detects various types of damage and identifies the areas to be enhanced. This comprehensive product removes scratches, snowflakes, and noise and upscales movies to modern standards. The improvement incorporates Super Resolution and Smart Enhancer capabilities to automate the clean-up process. Compared to a lengthy manual process that could take weeks, our solution can intelligently repair films in a few hours, drastically improving your productivity.

Increase the emotional impact of your videos on different screens by using our automatic region-of-interest cropping. The solution takes a horizontal screen video as input, detects the most significant area, crops the video to the user-specified aspect ratio, and then outputs the ideal vertical screen version of the video for mobile devices. This solution is perfect for re-purposing your content in the most impactful way on different screen sizes and increasing your distribution throughput of mobile content.

Our Smart Clipping solution analyzes full-length videos and identifies actions, events, and faces. The solution extracts each of these detections into individual clips in collections for fast retrieval for editing or publishing.

The Smart Highlight solution works with the Smart Clipping solution to automate the creation of exciting highlight videos. The highlight generator selects the clips for the video using parameters set by the user, such as showing all goal events in the football match. The algorithm can also automate the editing of the video highlights with a soundtrack library for complete videos ready for publishing.

The Smart Cover solution uses AI machine learning to find the most enticing and compelling frames to use as video or image thumbnails to attract viewers and increase the clickthrough rate.

Video Deduplication is used in video checking scenarios to identify duplicate or redundant video content for users to take further action. Duplicated content is identifiable regardless of resolution scaling, image flipping, noise, or subtitles. The solution first extracts the visual information of the videos into mathematical representations. Comparing the representation data of two videos can quickly determine the similarity. Applications of video deduplication include freeing up storage space in archives or triggering content take-down on sites due to copyright or ownership infringement.

Green screen replacement (also known as "chroma keying") could be tricky along edges and fine details such as hair. Media Lab's solution uses proprietary algorithms for background replacements that look natural and pleasing.

Media Lab's Watermark Detection uses AI deep learning to recognize watermarks such as logos and graphics embedded in videos and the associated creators. Business applications include detecting and filtering videos to avoid copyright infringement or triggering a recommendation engine to promote related videos or ads.

User-generated content (UGC) short videos are more popular than ever, with millions of video clips uploaded to short-video platforms. Manually viewing and tagging the videos is far too time-consuming to be effective in practice. Media Lab's Video Tagging solution uses AI deep learning to automate content scanning and tag generation of short videos. These tags can be utilized for video recommendation and detection systems further down the video distribution chain.