Roblox Corporation

Various implementations relate to methods, computer-readable media, and to automatically insert a script in a virtual environment. In some implementations, a method includes receiving a natural language prompt from a user, the natural language prompt specifying one or more functional requirements for the script. The method further includes providing the natural language prompt, a default prompt, and context information to a large language model (LLM). The method further includes obtaining, as output of the LLM and in response to the providing, the script and an attachment location for the script, wherein the attachment location identifies a particular entity associated with the virtual environment. The method further includes inserting the script in the virtual environment at the attachment location, wherein the script is executable in the virtual environment after the inserting.

Some implementations relate to methods, systems, and computer readable media for generating code based on three-dimensional (3D) virtual assets. A 3D model of a 3D virtual object is received. A latent representation of the 3D model is generated using a pre-trained encoder, and encodes semantic and structural information of the 3D model. A text prompt and the latent representation of the 3D model are provided as input to a code generation model, where the text prompt includes a request to generate code with references to the 3D virtual object and where the latent representation of the 3D model serves as a conditioning input to the code generation model. The code is generated by the code generation model upon execution on a virtual platform or a client device, and causes the virtual platform or client device to perform one or more actions with reference to the 3D virtual object.

According to one aspect of the present disclosure, a method of training a deformation prediction model is provided. In some implementations, a method includes obtaining a neutral expression three-dimensional (3D) mesh and a set of facial action coding system (FACS) weights, wherein the set of FACS weights represent a target facial pose or a target facial expression. The method further includes obtaining a predicted 3D mesh from the deformation prediction model, wherein the predicted mesh is arranged to at least partially mimic the target facial pose or target facial expression, rendering a two-dimensional (2D) image from the predicted mesh, and adjusting the deformation prediction model based on one or more 2D loss functions, the one or more 2D loss functions being based on comparison of the 2D image with a groundtruth 2D image obtained from a pre-trained 2D animation model.

Some implementations relate to translating text within images in a virtual environment while preserving the original style and visual characteristics of the text. In some implementations, the method includes obtaining an original image that includes text in a source language; recognizing content of the text and generating translated text in a target language; determining a text region of the text in the original image; determining a style encoding for the text; generating a masked version and noisy version of the original image; providing the noisy version, masked version, and text region as direct inputs to a pre-trained diffusion model; providing the translated text and the style encoding as conditioning inputs to the diffusion model; and obtaining an output image including the translated text, where a visual style of the translated text in the output image is the same as the visual style of the text in the original image.

A user query, such as a user query processed by the online game system, is provided as input into a trained machine learning model. The machine learning model predicts candidate languages of the user query and outputs language confidence scores for the candidate languages. The user query is also matched with stored query data associated with game information in individual language databases for the respective candidate languages. A match scores may be determined based on a certainty of the respective response matches. The match scores and the language confidence scores may be weighted to determine a weighted score. The weighted scores of the response matches are applied to decide which game information retrieved from the identified database is used in forming a response of search results to the user.

A user prompt, such as a user prompt received by a client device and sent to an online game system, is provided into a trained large language model. The large language model identifies keywords corresponding to the user prompt. These keywords may be provided to a search engine that identifies corresponding object(s) to place in a virtual experience. The large language model further processes the user prompt to determine spatial placement information for the objects and places the objects accordingly. Subsequently, the system may iteratively receive more prompts and update the virtual experience based on the additional prompts. The placement may be facilitated using macros. The prompts may also affect other attributes of the objects. The knowledge built-into the LLM allows it to suggest which objects are relevant and what quantity and arrangements of these objects is consistent with a scene requested in the user prompt.

A user query, such as a user query processed by the online game system, is provided as input into a trained machine learning model. The machine learning model predicts candidate languages of the user query and outputs language confidence scores for the candidate languages. The user query is also matched with stored query data associated with game information in individual language databases for the respective candidate languages. A match scores may be determined based on a certainty of the respective response matches. The match scores and the language confidence scores may be weighted to determine a weighted score. The weighted scores of the response matches are applied to decide which game information retrieved from the identified database is used in forming a response of search results to the user.

A user query, such as a user query processed by the online game system, is provided as input into a trained machine learning model. The machine learning model predicts candidate languages of the user query and outputs language confidence scores for the candidate languages. The user query is also matched with stored query data associated with game information in individual language databases for the respective candidate languages. A match scores may be determined based on a certainty of the respective response matches. The match scores and the language confidence scores may be weighted to determine a weighted score. The weighted scores of the response matches are applied to decide which game information retrieved from the identified database is used in forming a response of search results to the user.

Implementations relate to systems, methods, and computer-readable media to generate text tags for games. A computer-implemented method is provided to generate one or more text tags for a game using a trained machine learning model. Data that includes a game identifier of the game and a set of digital assets associated with the game are provided as input to the trained machine learning model. Predicted text tags are generated using the trained machine learning model based on the set of digital assets associated with the game. The text tags are associated with a respective prediction score. One or more text tags are selected from the plurality of predicted text tags based on the respective prediction score.

A user query, such as a user query processed by the online game system, is provided as input into a trained machine learning model. The machine learning model predicts candidate languages of the user query and outputs language confidence scores for the candidate languages. The user query is also matched with stored query data associated with game information in individual language databases for the respective candidate languages. A match scores may be determined based on a certainty of the respective response matches. The match scores and the language confidence scores may be weighted to determine a weighted score. The weighted scores of the response matches are applied to decide which game information retrieved from the identified database is used in forming a response of search results to the user.