Granted Patent B2

US 12,646,266 · App. 18/497,629 · Granted Jun 2, 2026

Inventors: Daria Skrypnyk (Los Angeles, CA); Matthew Hallberg (Los Angeles, CA)

Assignee: SNAP INC.

G06T19/006G06T15/04G06T17/20G06T19/20G10L15/1815G10L15/22G06T2219/2004G06T2219/2012G10L2015/223

Abstract

Claims (42)

1 . A system comprising:

at least one processor;

at least one memory component storing instructions that, when executed by the at least one processor, cause the at least one processor to perform operations comprising:

identifying a prompt of a user indicating a user's intent;

accessing an image template, wherein the image template includes placement of features within the image template;

processing a combination of data associated with the image template and the prompt using a generative machine learning model to generate a first populated image template in which one or more portions of the image template are populated with visual content representing the user's intent;

accessing an image depicting a real-world object, wherein the real-world object is of the user's body, wherein the image template includes placement of a head, one or more limbs, and a torso; and

overlaying the first populated image template comprising the visual content representative of the user's intent on at least a portion of the real-world object based on the placement of the features of the image template.

2 . The system of claim 1 , wherein overlaying the first populated image template on at least a portion of the real-world object comprises adjusting a 3D mesh for the real-world object based on the first populated image template.

3 . The system of claim 2 , wherein the operations further comprise receiving a second populated image template from the generative machine learning model, wherein adjusting the 3D mesh is based on the first and second populated image templates.

4 . The system of claim 3 , wherein the first populated image template comprises depth information, wherein the second populated image template comprises color.

5 . The system of claim 4 , wherein the real-world object is of a face of the user, wherein the image template includes placement of facial features, wherein the 3D mesh is a mesh for a human face.

6 . The system of claim 4 , wherein adjusting the 3D mesh comprises generating a point cloud of 3D points representing spatial information by mapping pixel coordinates of the second populated image template with the pixel coordinates of the depth information in the first populated image template.

7 . The system of claim 1 , wherein the features include at least two eyes, a nose, and a mouth, wherein the first populated image template includes depth information for one or more of the features within the first populated image template.

8 . The system of claim 1 , wherein the generative machine learning model includes a stable diffusion model.

9 . The system of claim 1 , wherein identifying the prompt comprises receiving a voice or text command from the user.

10 . The system of claim 1 , wherein identifying the prompt comprises determining an intent of the user based on environmental characteristics identified from data received from an interaction client of the user.

11 . The system of claim 10 , wherein identifying the prompt comprises determining the intent of the user based on multimodal memory embeddings associated with the user.

12 . The system of claim 1 , wherein the first populated image template comprises an augmented image of the image template based on one or more characteristics indicated in the prompt.

13 . The system of claim 1 , wherein the generative machine learning model is trained to generate populated image templates of peoples' faces based on a structure of human face features placed within image templates, the populated image template being populated based on an artificial texture generated by the generative machine learning model using the prompt.

14 . The system of claim 1 , wherein the operations further comprise training the generative machine learning model by:

collecting a dataset of images related to real-world objects;

collecting textual descriptions or prompts for each real-world object in the dataset of real-world objects that describe one or more characteristics within a corresponding image;

inputting the dataset of images and textual descriptions or prompts to the generative machine learning model;

using a denoising score matching objective to determine a loss function; and

updating one or more parameters in the generative machine learning model to reduce the loss function.

15 . The system of claim 1 , wherein the operations further comprise continuously overlaying the first populated image template on the portion of the real-world object in response to continuous capture of a camera feed from a camera system.

16 . The system of claim 1 , wherein the prompt includes an indication of the intent for a particular animal, wherein real-world object is of the user's face.

17 . The system of claim 16 , wherein the first populated image template includes the particular animal modified based on a structure of features within the image template.

18 . A method comprising:

identifying a prompt of a user indicating a user's intent;

accessing an image template, wherein the image template includes placement of features within the image template;

processing a combination of data associated with the image template and the prompt using a generative machine learning model to generate a first populated image template in which one or more portions of the image template are populated with visual content representing the user's intent;

receiving a second populated image template from the generative machine learning model;

accessing an image depicting a real-world object; and

overlaying the first populated image template comprising the visual content representative of the user's intent on at least a portion of the real-world object based on the placement of the features of the image template, overlaying the first populated image template by adjusting a 3D mesh.

19 . A non-transitory computer-readable storage medium storing instructions that, when executed by at least one processor, cause the at least one processor to perform operations comprising:

identifying a prompt of a user indicating a user's intent based on multimodal memory embeddings associated with the user;

accessing an image template, wherein the image template includes placement of features within the image template;

processing a combination of data associated with the image template and the prompt using a generative machine learning model to generate a first populated image template in which one or more portions of the image template are populated with visual content representing the user's intent;

accessing an image depicting a real-world object; and

overlaying the first populated image template comprising the visual content representative of the user's intent on at least a portion of the real-world object based on the placement of the features of the image template.

Assignments (1)

Continuity (2)

References Cited (400)