Abstract: This specification describes a computer-implemented method of generating speech audio for use in a video game, wherein the speech audio is generated using a voice convertor that has been trained to convert audio data for a source speaker into audio data for a target speaker. The method comprises receiving: (i) source speech audio, and (ii) a target speaker identifier. The source speech audio comprises speech content in the voice of a source speaker. Source acoustic features are determined for the source speech audio. A target speaker embedding associated with the target speaker identifier is generated as output of a speaker encoder of the voice convertor. The target speaker embedding and the source acoustic features are inputted into an acoustic feature encoder of the voice convertor. One or more acoustic feature encodings are generated as output of the acoustic feature encoder. The one or more acoustic feature encodings are derived from the target speaker embedding and the source acoustic features.

Abstract: Embodiments of the systems and methods disclosed herein provide a sponsor matching system in which players and sponsors can be matched. Upon a match based at least in part on stored sponsorship criteria and/or player preferences, a first sponsor can select a set of players to receive permission to select an advertisement associated with the first sponsor. Once a first player of the selected players selects an advertisement and an advertisement placement location associated with the first sponsor, the sponsor matching system can generate game rendering instructions for a first player system associated with the first player.

Abstract: A computer-implemented method comprises instantiating a policy function approximator. The policy function approximator is configured to calculate a plurality of estimated action probabilities in dependence on a given state of the environment. Each of the plurality of estimated action probabilities corresponds to a respective one of a plurality of discrete actions performable by the reinforcement learning agent within the environment. An initial plurality of estimated action probabilities in dependence on a first state of the environment are calculated. Two or more of the plurality of discrete actions are concurrently performed within the environment when the environment is in the first state. In response to the concurrent performance, a reward value is received. In response to the received reward value being greater than a baseline reward value, the policy function approximator is updated, such that it is configured to calculate an updated plurality of estimated action probabilities.

Abstract: A system and method for facilitating a user defined virtual space is disclosed. One or more virtual space locations and/or activities may be correlated with user specified geolocations. In some implementations, the user specified geolocations may be verified against one or more spatial requirements prior to recording the user selected space-geolocation correlations. A user request to initiate an action or activity in the virtual space may be received. Prior to executing the requested action or activity in the virtual space, the user current geolocation may be verified against that specified in a space-geolocation correlation corresponding to a virtual space location or activity indicated in the user request.

Abstract: Various aspects of the subject technology relate to systems, methods, and machine-readable media for learning a foreign language. The method includes executing a video game in a first human language. The method includes pausing gameplay of the video game for a paused time instance. The method includes executing a digital mini-puzzle game during the paused time instance in the gameplay of the video game, the digital mini-puzzle game executed in a second human language, the digital mini-puzzle game executed utilizing assets of the video game. The method includes receiving a response to the digital mini-puzzle game from a player-computing device corresponding to a player, the response comprising at least one of the first human language and/or the second human language. The method includes determining a score of the response corresponding to the player based at least in part on a comparison of the response with translation pairs in a database.

Abstract: A visual-programming tool processes nodes of a graph corresponding to operations or functions in program code associated with a plurality of programs, (e.g., games), stored as graph of nodes with logical connections signifying inputs, outputs, and/or units of connected nodes. The visual-programming tool resolves valid types and/or units associated with respective connected nodes and can propagate valid types and/or units throughout the graph.

Abstract: Systems and methods for conservation of bandwidth and improved user experience via enhanced streaming of video games. An example method includes receiving a request to remotely play a video game, the video game being executed by the system and streamed to a user device for presentation. The video game is executed, and rendered image frames are generated. Geometry data associated with the rendered image frames is generated, with the geometry data representing locations of geometric elements that form geometry utilized, by the video game, to generate the rendered image frames. The rendered image frames are encoded into a gameplay stream. A first stream comprising the encoded gameplay stream and a second stream comprising the geometry data are provided to the user device. The user device is configured to perform post-processing effects on the rendered image frames encoded in the gameplay stream prior to display on the user device.

Abstract: A computer-implemented method is provided of allowing a user to automatically transform domain knowledge into a machine learning model to be used in real-time operation of video games. The method comprises providing a user interface which allows a user to define domain knowledge relating to a video game by specifying one or more labeling functions; transforming the labeling functions into executable code; labeling raw data relating to the video game using the executable code to obtain labeled data; and applying an automated machine learning module to the labeled data to obtain a machine learning model.

Abstract: A computer-implemented method comprises receiving a first compressed representation of a texture map in a first compression format, wherein the first compressed representation has been compressed using a first compressor, and receiving an array of compression parameters for a second compressor, the array of compression parameters including one or more respective compression parameters for each of a plurality of pixel regions of the texture map. The method further comprises decompressing the first compressed representation of the texture map to obtain the texture map, and compressing, using the second compressor, the texture map to a second compressed representation in a second compression format, comprising compressing each of said plurality of pixel regions of the texture map in accordance with the respective one or more compression parameters.

Abstract: A system may render glyphs based on stored textures without loss of quality at subpixel scales. The system may determine a content of a pixel of a display corresponds to a glyph, determine a subpixel alignment offset of a specified screen coordinates for the glyph with respect to the pixels of the display, based on the subpixel alignment offset, select one or more versions of the glyph from a plurality of versions of the glyph, a first version of the glyph of the plurality of versions of the glyph having a corresponding first subpixel alignment offset and a second version of the glyph of the plurality of versions of the glyph having a corresponding second subpixel alignment offset, and generate a display version of the pixel based on the selected one or more versions of the glyph and the subpixel alignment offset of the specified screen coordinates.

Abstract: A computer-implemented method for generating a machine-learned model to generate facial position data based on audio data comprising training a conditional variational autoencoder having an encoder and decoder. The training comprises receiving a set of training data items, each training data item comprising a facial position descriptor and an audio descriptor; processing one or more of the training data items using the encoder to obtain distribution parameters; sampling a latent vector from a latent space distribution based on the distribution parameters; processing the latent vector and the audio descriptor using the decoder to obtain a facial position output; calculating a loss value based at least in part on a comparison of the facial position output and the facial position descriptor of at least one of the one or more training data items; and updating parameters of the conditional variational autoencoder based at least in part on the calculated loss value.

Abstract: Systems and methods are described herein for monitoring a gameplay session for violations of a policy and creating a remediation gameplay session through which remediation can be provided to players or player accounts that violate gameplay policies. The systems and methods can create a remediation gameplay session based in part on the game state data of the gameplay session during which the violation occurs.

Abstract: Systems and methods are provided for enhanced animation generation based on using motion mapping with local bone phases. An example method includes accessing first animation control information generated for a first frame of an electronic game including local bone phases representing phase information associated with contacts of a plurality of rigid bodies of an in-game character with an in-game environment. Executing a local motion matching process for each of the plurality of local bone phases and generating a second pose of the character model based on the plurality of matched local poses for a second frame of the electronic game.

Abstract: A system may provide a programming environment combining visual programming and text based programming. The system may determine a first node and a second node of a first plurality of nodes are native function call nodes of a visual graph-based programming language and determine a third node of the first plurality of nodes is a text-based node including text-based statements in a text-based programming language. In response to determining the third node of the first plurality of nodes is the text-based node, the system may interpret the text-based statements into a second plurality of nodes, the second plurality of nodes being native function call nodes of the visual graph-based programming language. The system may then generate a combined code based on the first plurality of nodes and the second plurality of nodes and compile the combined code.

Abstract: A method for hair rasterization includes receiving hair data of hair strands, each hair strand including one or more hair segments; determining a first set of hair segments that is not fully occluded by one or more opaque objects; classifying each hair segment in the first set of hair segments into buckets corresponding to one or more tiles of pixels that the hair segment overlaps and one or more clusters arranged by depth per tile; for each tile, rasterizing the hair segments included in a first cluster for the tile, the first cluster for the tile being the cluster that is nearest to a location of a camera that includes at least one hair segment, and accumulating color and opacity for each pixel of the tile based on rasterizing the hair segments included in the first cluster; and combining the tiles that include rasterized hair segments into a hair overlay.

Abstract: In the disclosed systems and methods for competitive scene completion, in conjunction with a scene completion challenge, an image of an initial scene and a plurality of markers are displayed. For each user marker selection, virtual furnishing units corresponding to the unit type are displayed. User unit selection results in display of a three-dimensional graphic of the selected virtual furnishing unit at the corresponding coordinates within the scene, thereby creating an augmented scene that comprises the initial scene with three-dimensional graphics of selected virtual furnishing units. The augmented scene is submitted to a remote server. The user is provided with a reward that consists of credits. Responsive to user selection to access the store, a user interface for the store is displayed within the application. Visual representations of tangible products are displayed. The credits are configured for use towards purchase of the tangible products.

Abstract: In a video game, a player's character can start in a normal state, receive first damage, and change to an incapacitated state. The player's character can be revived from the incapacitated state back to the normal state. The player's character can be changed from the incapacitated state to a preliminarily defeated state, and in response, a player respawn activation item can be generated. The player respawn activation item can be used by the player's teammates to respawn the player's character at one or more respawn locations.

Abstract: According to a first aspect of this specification, there is disclosed a computer implemented method comprising: training, based on an initial behavior goal and using reinforcement-learning, a reinforcement-learning model for controlling behavior of a non-playable character in a computer game environment; converting the trained reinforcement-learning model into a behavior tree model for controlling behavior of the non-playable character; editing, based on a user input, the behavior tree model to generate an updated behavior tree model for controlling behavior of the non-playable character; and outputting a final model for controlling non-player character behavior for use in the computer game environment, wherein the model for controlling non-player character behavior is based at least in part on the updated behavior tree model.

Abstract: Systems and methods for enhanced training of machine learning systems based on automatically generated visually realistic gameplay. An example method includes obtaining electronic game data that includes rendered images and associated annotation information, the annotation information identifying features included in the rendered images to be learned, and the electronic game data being generated by a video game associated with a particular sport. Machine learning models are trained based on the obtained electronic game data, with training including causing the machine learning models to output annotation information based on associated input of a rendered image. Real-world gameplay data is obtained, with the real-world gameplay data being images of real-world gameplay of the particular sport. The obtained real-world gameplay data is analyzed based on the trained machine learning models. Analyzing includes extracting features from the real-world gameplay data using the machine learning models.

Abstract: A game management system identifies gaming data associated with online game players and determines, based at least in part on the gaming data, other players to recommend as friends for playing an online game. Gaming data from online games other than the online game for which friend recommendations are to be made may be used to provide the friend recommendations. A subset of categories of the gaming data may be used to initially bin the plurality of players into separate bins. Similarity of players, with respect to their gaming data, or a subset thereof, within each bin may be determined. This similarity analysis may result in a similarity score corresponding to each pair of players within a bin. The similarity scores may be used to determine if two players are compatible from an online gaming standpoint and provide friend recommendations to players.