UnfoldIPThemes

Company patents

Agora Lab, Inc.

Agora Lab, Inc. shows a surprising shift in its patent strategy, with a significant decline in its core Streaming & Real-Time Media category, which saw a -100.0% YoY drop in 2026 (so far in 2026) and now holds only 0 patents, despite being 26.5% of its portfolio. Concurrently, Speech Processing is emerging as a key focus, experiencing a robust +33.3% YoY growth in 2026 (so far in 2026) and now representing 16.2% of the portfolio.

Patent Trend by Technology Area

Yearly patent publications since 2023

Product themes

Product-level themes inferred from filings since 2023, with category chips showing where each theme appears. Select a theme to filter the patents below.

68 US filings (since 2023) · 12 categories · 17 themes

Video Quality & Encoding Optimization

Methods and apparatus for improving the visual fidelity, resolution, or compression efficiency of video signals, often through advanced processing, up-scaling, or neural network-based filters.

Pictorial / Video CommunicationsComputer Vision
Who else files here? →
24since 2023
+66.7%YoY
Video Enhancement & Object Tracking

Methods and systems for improving the quality of video streams, generating intermediate frames, or continuously locating and following objects within a sequence of images, even under occlusion.

Image Processing
Who else files here? →
22since 2023
+66.7%YoY
Voice & Audio Quality Enhancement

Techniques for improving the perceived quality, synchronization, and moderation of audio and voice streams, often involving codec management, transcoding, and content analysis.

Streaming & Real-Time Media
Who else files here? →
12since 2023
+50.0%YoY
Communication Network Session Management

Systems and methods for establishing, maintaining, modifying, and terminating communication sessions across various network architectures, including service discovery, resource allocation, and resilience mechanisms.

Streaming & Real-Time Media
Who else files here? →
10since 2023
-33.3%YoY
Spatial Audio Encoding & Compression

Techniques for efficiently representing, compressing, and decompressing spatial audio information, such as inter-channel phase/time differences, directional components, or higher-order ambisonics, for storage or transmission.

Stereophonic Audio
Who else files here? →
8since 2023
-33.3%YoY
XR/Virtual Environment Audio

Systems and methods for integrating and managing spatial audio within extended reality (XR), virtual reality (VR), or augmented reality (AR) environments, often involving dynamic audio source positioning, streaming, and synchronization with visual elements.

Stereophonic Audio
Who else files here? →
8since 2023
-33.3%YoY
Advanced Audio Signal Processing

Techniques for enhancing, encoding, decoding, or separating speech and audio signals, often involving multi-microphone arrays, acoustic echo cancellation, beamforming, or advanced audio compression for improved clarity and quality.

Speech Processing
Who else files here? →
8since 2023
new
Network Operations, Reliability & Traffic

Techniques and systems designed to monitor network health, diagnose issues, optimize traffic flow, and ensure continuous operation and reduced downtime in complex network environments, including cloud and storage area networks.

Network Management & MonitoringRouting, Switching & QoS
Who else files here? →
7since 2023
0.0%YoY
Network Traffic Management & QoS

Techniques and systems for optimizing network traffic flow, distributing loads across multiple paths or resources, and ensuring quality of service based on various criteria like application type, latency, or resource availability. This includes dynamic path selection, congestion control, and resource allocation.

Routing, Switching & QoS
Who else files here? →
7since 2023
0.0%YoY
Large Model Text Generationfiltered

Techniques for generating human-like text or other content using large pre-trained models, often involving prompt engineering, speculative decoding, or multi-modal inputs for content creation.

Machine Learning & AI
Who else files here? →
6since 2023
-50.0%YoY
Generative AI for Images

Techniques utilizing deep learning models like Generative Adversarial Networks (GANs) or diffusion models to create new images, modify existing ones, or generate synthetic data based on various inputs or conditions.

Computer VisionImage Processing
Who else files here? →
6since 2023
-50.0%YoY
Adaptive & Personalized Audio

Adjusting audio presentation parameters in real-time or based on user-specific profiles, environmental conditions, or content characteristics to optimize the listening experience.

Stereophonic Audio
Who else files here? →
6since 2023
-66.7%YoY
Adaptive Media Delivery & Caching

Methods and systems for efficiently distributing and delivering media content, including techniques for multi-source streaming, content caching, and optimizing delivery based on network conditions or device capabilities.

Streaming & Real-Time Media
Who else files here? →
5since 2023
+100.0%YoY
Real-time Multimedia Conferencing

Technologies enabling synchronous, interactive multimedia communication sessions, including user interfaces, content sharing, and underlying session management for multiple participants.

Streaming & Real-Time Media
Who else files here? →
5since 2023
n/a
Context-Aware Conversational AI

AI systems designed to engage in natural language dialogue, maintain conversation state, understand user intent, and generate relevant responses, often across multiple communication channels or modalities.

Speech Processing
Who else files here? →
1since 2023
new
Adaptive Speech Synthesis & Messaging

Technologies for generating artificial speech that is personalized, context-aware, or adaptable to specific virtual agents or messaging campaigns, often utilizing text-to-speech (TTS) and audio caching for efficient delivery.

Speech Processing
Who else files here? →
1since 2023
new
Deepfake Voice Detection

Methods and systems for identifying synthetic or manipulated speech (deepfake audio) using forensic analysis of audio features, such as breath patterns, vocoder signatures, or machine learning models to determine authenticity.

Speech Processing
Who else files here? →
1since 2023
n/a

Patents

Showing 1-9 of 9

Large Model Text Generation
Page 1 of 1
US 12314862 B2GRANTED
G06N3/084

Optimizing generative networks via latent space regularizations

Filed:2024-04-26Pub:2025-05-27
Applicant:Agora Lab, Inc.

A method for image generation based on a Generative AI Network. The Generative AI Network includes a generator and an encoder. The method includes determining, by the encoder, a first encoding E(Y) of a target image Y; generating, by the generator, a generated image G(Z) corresponding to the target image Y, wherein the generated image G(Z) is located in a close vicinity of a target neighborhood of the target image Y, and outputs of the generator are mapped, by the encoder, to a latent space adaptable to manipulate at least one characteristics of images generated by the Generative AI Network; and generating, by the encoder, a second encoding E(G(Z)) of the generated image G(Z) corresponding to the target image Y, wherein the first and second encodings E(Y) and E(G(Z)) map the target image Y and the generated image G(Z) to the latent space.

US 20240296332 A1APPLICATION
G06N3/084

Optimizing Generative Networks Via Latent Space Regularizations

Filed:2024-04-26Pub:2024-09-05
Applicant:Agora Lab, Inc.

A method for image generation based on a Generative AI Network. The Generative AI Network includes a generator and an encoder. The method includes determining, by the encoder, a first encoding E(Y) of a target image Y; generating, by the generator, a generated image G(Z) corresponding to the target image Y, wherein the generated image G(Z) is located in a close vicinity of a target neighborhood of the target image Y, and outputs of the generator are mapped, by the encoder, to a latent space adaptable to manipulate at least one characteristics of images generated by the Generative AI Network; and generating, by the encoder, a second encoding E(G(Z)) of the generated image G(Z) corresponding to the target image Y, wherein the first and second encodings E(Y) and E(G(Z)) map the target image Y and the generated image G(Z) to the latent space.

US 12001956 B2GRANTED
G06N3/084

Optimizing supervised generative adversarial networks via latent space regularizations

Filed:2023-05-17Pub:2024-06-04
Applicant:Agora Lab, Inc.

A method for image generation based on a Generative Adversarial Network (GAN) including a generator, a discriminator, and an encoder, wherein outputs of the generator are mapped, by the encoder, to a latent space adaptable to manipulate at least one characteristics of images generated by the GAN, the method including generating, by the encoder, a first encoding E(Y) of a target image Y and a second encoding E(G(Z)) of a generated image G(Z) corresponding to the target image Y, wherein the first and second encodings E(Y) and E(G(Z)) map Y and G(Z) to the latent space having a lower dimension than dimensionality of Y and G(Z), wherein the encoder is trained to minimize the differences between the first and second encodings E(Y) and E(G(Z)), and the generator is trained by using the first and second encodings E(Y) and E(G(Z)) as part of a loss function.

US 20230289608 A1APPLICATION
G06N3/084

Optimizing Supervised Generative Adversarial Networks via Latent Space Regularizations

Filed:2023-05-17Pub:2023-09-14
Applicant:Agora Lab, Inc.

A method for image generation based on a Generative Adversarial Network (GAN) including a generator, a discriminator, and an encoder, wherein outputs of the generator are mapped, by the encoder, to a latent space adaptable to manipulate at least one characteristics of images generated by the GAN, the method including generating, by the encoder, a first encoding E(Y) of a target image Y and a second encoding E(G(Z)) of a generated image G(Z) corresponding to the target image Y, wherein the first and second encodings E(Y) and E(G(Z)) map Y and G(Z) to the latent space having a lower dimension than dimensionality of Y and G(Z), wherein the encoder is trained to minimize the differences between the first and second encodings E(Y) and E(G(Z)), and the generator is trained by using the first and second encodings E(Y) and E(G(Z)) as part of a loss function.

US 11694085 B2GRANTED
G06K9/62

Optimizing supervised generative adversarial networks via latent space regularizations

Filed:2021-05-19Pub:2023-07-04
Applicant:Agora Lab, Inc.

A method of training a generator G of a Generative Adversarial Network (GAN) includes receiving, by an encoder E, a target data Y; receiving, by the encoder E, an output G(Z) of the generator G, where the generator G generates the output G(Z) in response to receiving a random sample Z and where a discriminator D of the GAN is trained to distinguish which of the G(Z) and the target data Y; training the encoder E to minimize a difference between a first latent space representation E(G(Z)) of the output G(Z) and a second latent space representation E(Y) of the target data Y, where the output G(Z) and the target data Y are input to the encoder E; and using the first latent space representation E(G(Z)) and the second latent space representation E(Y) to constrain the training of the generator G.

US 11544571 B2GRANTED
G06N20/20

Optimizing unsupervised generative adversarial networks via latent space regularizations

Filed:2019-10-23Pub:2023-01-03
Applicant:Agora Lab, Inc.

Training a generator G of a GAN includes generating, by G and in response to receiving a first input Z, a first output G(Z); generating, by an encoder E of the GAN and in response to receiving the first output G(Z) as input, a second output E(G(Z)); generating, by G and in response to receiving the second output E(G(Z)) as input, a third output G(E(G(Z))); generating, by E and in response to receiving the third output G(E(G(Z))) as input, a fourth output E(G(E(G(Z)))); training E to minimize a difference between the second output E(G(Z)) and the fourth output E(G(E(G(Z)))); and using the second output E(G(Z)) and fourth output E(G(E(G(Z)))) to constrain a training of the generator G. G(Z) is an ambient space representation Z. E(G(Z)) is a latent space representation of G(Z). G(E(G(Z))) is an ambient space representation of E(G(Z)). E(G(E(G(Z)))) is a latent space representation of G(E(G(Z))).

US 20210271933 A1APPLICATION
G06K9/62

Optimizing Supervised Generative Adversarial Networks via Latent Space Regularizations

Filed:2021-05-19Pub:2021-09-02
Applicant:Agora Lab, Inc.

A method of training a generator G of a Generative Adversarial Network (GAN) includes receiving, by an encoder E, a target data Y; receiving, by the encoder E, an output G(Z) of the generator G, where the generator G generates the output G(Z) in response to receiving a random sample Z and where a discriminator D of the GAN is trained to distinguish which of the G(Z) and the target data Y; training the encoder E to minimize a difference between a first latent space representation E(G(Z)) of the output G(Z) and a second latent space representation E(Y) of the target data Y, where the output G(Z) and the target data Y are input to the encoder E; and using the first latent space representation E(G(Z)) and the second latent space representation E(Y) to constrain the training of the generator G.

US 11048980 B2GRANTED
G06K9/62

Optimizing supervised generative adversarial networks via latent space regularizations

Filed:2019-08-02Pub:2021-06-29
Applicant:Agora Lab, Inc.

A method of training a generator G of a Generative Adversarial Network (GAN) includes receiving, by an encoder E, a target data Y; receiving, by the encoder E, an output G(Z) of the generator G, where the generator G generates the output G(Z) in response to receiving a random sample Z that is a noisy sample, and where a discriminator D of the GAN is trained to distinguish which of the G(Z) and the target data Y is real data; training the encoder E to minimize a difference between a first latent space representation E(G(Z)) of the output G(Z) and a second latent space representation E(Y) of the target data Y, where the output G(Z) and the target data Y are input to the encoder E; and using the first latent space representation E(G(Z)) and the second latent space representation E(Y) to constrain the training of the generator G.

US 11048974 B2GRANTED
G06K9/62

Effective structure keeping for generative adversarial networks for single image super resolution

Filed:2019-08-05Pub:2021-06-29
Applicant:Agora Lab, Inc.

A method of training a generator G of a Generative Adversarial Network (GAN) includes generating a real contextual data set {x 1 , . . . , x N } for a high resolution image Y; generating a generated contextual data set {g 1 , . . . , g N } for a generated high resolution image G(Z); calculating a perceptual loss L pcept value using the real contextual data set {x 1 , . . . , x N } and the generated contextual data set {g 1 , . . . , g N }; and training the generator G using the perceptual loss L pcept value. The generated high resolution image G(Z) is generated by the generator G of the GAN in response to receiving an input Z, where the input Z is a random sample that corresponds to the high resolution image Y.

1