
Attention Labs, an engagement-control layer for voice AI, whose SAA performs real-time addressee detection before audio reaches the ASR.
attentionlabs.aiCrawled Nov 3, 2026
Addressee detection decides if speech is directed at the device, before ASR runs. SAA makes this call in real time, no wake word required.
Attention Labs published research on addressee detection for voice AI, available on arXiv. A free evaluation is offered.
Drive-thru AI fires on backseat, radio, and crew. SAA adds pre-ASR addressee detection, routing only the ordering driver's voice before ASR runs. Free eval.
Users can test addressee detection on their own audio, receiving scored results within 48 to 72 hours.
The page explains how diarization identifies speakers and timing, while addressee detection determines if speech was directed at the device before automatic speech recognition.
A wake word is a user-performed protocol that still misfires. Addressee detection classifies device-directed speech before ASR, no trigger, fails closed.
The cocktail party problem for voice AI has two halves. SAA adds addressee detection before ASR, in real time, routing only device-directed speech onward.
SAA canonical facts: pre-ASR addressee detection in real time, before speech-to-text (arXiv:2604.08412), per-device licensing, and a free eval tier.
Attention Labs seeks an Account Executive for enterprise and OEM device deals in San Francisco.
Attention Labs hires for roles building an engagement control layer for people and AI.
Attention Labs is hiring a Developer Advocate to support its SDK, including documentation and developer feedback.
Attention Labs offers a role for an Embedded Systems Engineer to deploy addressee detection models on OEM hardware in Memphis.
Attention Labs develops an engagement control system enabling true telepresence for people and AI across devices and models.
The live demo shows addressee detection responding to direct speech while ignoring nearby conversations.
SAA by Attention Labs determines if speech is addressed to an AI before any model activates, available for on-device licensing.
Data Processing Addendum for attention labs enterprise and commercial customers subject to GDPR, UK GDPR, or equivalent data protection obligations.
Privacy Policy for attention labs products and services, including the SAA Cloud API.
Current list of sub-processors engaged by Socero Inc. to operate the SAA Cloud API.
Acceptable Use Policy for the attention labs website, SAA Cloud API, Paid POC program, and developer SDKs.
The Security and Trust Center covers attention Labs’ security posture and responsible disclosure process for the SAA Cloud API.
Terms of Service governing use of the attention labs website, SAA Cloud API, and related developer services.
Attention Labs introduces **SAA**, a software layer that filters voice commands for AI agents. The system determines whether spoken input is directed at a device before passing it to the AI stack, enabling responses only when explicitly addressed. This reduces unintended activations and improves targeted interactions for voice-controlled systems.
Jun 2026The National Law ReviewNews article
Attention Labs receives the **CES Picks Award 2026** for its AI system designed to engage in group conversations. The technology intelligently determines when to listen, speak, or remain silent during multi-party discussions. The Toronto-based startup showcased the innovation at **CES 2026**, highlighting its potential for naturalistic AI interaction.
Jan 2026Big News Network.comNews article
Attention Labs demonstrates how conversational AI struggles in shared environments like offices and vehicles. Current voice systems lack the ability to distinguish between users or determine when responses are relevant, despite widespread adoption in homes and workplaces. The demo underscores the need for AI that adapts to multi-person interactions rather than defaulting to single-user responses.
Jan 2026NERDBOTNews article
Attention Labs secures a **CES 2026 Picks Award** for its **selective attention AI**, which manages real-time group conversations by processing multi-talker interactions and determining optimal speaking moments. The company demonstrated this technology live at CES 2026, showcasing unscripted, natural multi-person dialogue handling. The award highlights its innovation in AI-driven conversational dynamics for machines.
Jan 2026financialcontent.comNews article