Abstract: This disclosure relates to a system, method, and computer program for facilitating video creation by recommending next steps in a user interface for video creation. A video production system tracks the state of a video as a user makes changes to the video. As the user develops the video, the system predicts the next step in the video generation process and makes recommendations to the user based on this prediction. The system will recommend an asset type to add to the video and also suggest specific instances of the asset type to add to the video. The video production system leverages a number of prediction models. The models include a deep learning model that is trained on a large corpus of video material to predict a next step of a video based on a current state. The system greatly shortens the time needed for the production of digital video by recommending video assets automatically to the user.

Abstract: This disclosure relates to a system, method, and computer program for automatically generating audio tracks for a video based on a current state of the video. The system provides a novel way to produce one or more audio tracks for a video. The system enables a user to enter a request for an audio track. In response to receiving the request, the system identifies the current state of a video, including the assets, scenes, and timelines in the video. Attributes of the current state of the video are then used to guide the output of machine-learning audio-generation models, which generate one or more audio clips for the video. In this way, visual assets added to a video can be used to guide the output of machine learning models that generate audio for the video. The sound clips produced by the machine-learning modules are then mixed to produce an audio track for the video.

Abstract: This disclosure relates to a system, method, and computer program for enabling an interactive process for video generation in which a user is able to guide the output of machine-learning asset enhancement modules to produce assets for a video. The system enables the user to interact with the asset enhancement modules via proxy elements in a video production workspace. The system provides a novel way to produce and edit video. The system enables any asset added to a video production workspace to serve as a proxy asset that a user can leverage to guide the output of one or more machine learning models trained to generate a type of multimedia asset. Specifically, a user can select an asset in the video production workspace, provide input on what attributes the user would like the asset to have, and the system then uses one or more machine-learning modules to generate an asset with the attributes requested by the user.

Abstract: The present disclosure describes a system, method, and computer program for encoding and decoding a unique signature for a user in a video file, wherein the digital signature is substantially invisible to the viewer of the video, easy to detect during the coding process, does not require specialized equipment for playback of the video, and is resilient to transcoding, video editing, and high compression encoding. A numeric digital signature is converted into a binary number. The binary number is then divided into sequences, where sequence includes a subset of data bits from the binary number. Each sequence is also associated with a sequence ID that enables the data bits in the sequence to be later combined in the correct order to recreate the binary number. A watermark is created for each of the sequences and embedded within suitable areas of a color channel.

Abstract: The present disclosure describes a system, method, and computer program for encoding and decoding a unique signature for a user in a video file, wherein the video file was created using a video format that does not specifically support embedding a unique signature in the video file. A unique signature, comprising a plurality of data bits, is associated with a user and divided into groups. For each group, a different sine wave is created for each bit in the group. The frequencies of the sine waves correspond to the types of bits, and the amplitudes of the sine waves indicate the values of the bits. Group signals are overdubbed on the infrasound range of the audio track of the video file. The unique signature is decoded from the video file by analyzing the frequencies and amplitudes of sine waves in the infrasound range of the audio track.