Meta launched real-time video chat capabilities for its Muse AI agent on November 5, 2026, enabling users across the United States to engage in face-to-face conversations with their digital avatars. The update, available on mobile, web, and Mac, adds a visual interface to the personal AI assistant originally introduced on September 8, 2026.
Meta Adds Video Chat to Muse
The rollout follows a preview of the technology at Meta’s Connect conference on September 23, 2026. Users can now interact with their AI agent through spoken language rather than relying solely on text-based prompts.
The system utilizes Muse Realtime Voice to provide immediate responses, allowing users to interrupt the agent during a conversation. This new functionality adds speech and visual presence to the established model where the agent handles tasks and requests approvals.
The digital avatar, which users can customize to represent various characters or personas, is designed to react dynamically to the user. Users can choose to keep interactions voice-only, place the avatar beside the conversation, or switch to a fullscreen face-to-face view.
While the default avatar is a cute, animated character nicknamed Jolly, users can change its appearance to look like an animal, a person, or another character. Meta stated that the avatar will use subtle expressions and gestures
that correspond to the tone of the interaction.
The system is optimized for real-time performance, producing 448-by-768 portrait video at 25 frames per second. It achieves approximately 870 milliseconds of latency from the end of a user’s turn to the first byte of the synchronized voice-and-video response. To make that process usable in real time, Meta employed distillation techniques, reducing generation from 120 guided teacher evaluations per chunk to two unguided student evaluations.
Beyond simple queries, Muse can manage delegated actions such as booking travel or sending emails. The agent can continue working after the app has been closed. It will return to the user when a sensitive action, such as a purchase, requires approval or when a specific change occurs.
Meta notes that hundreds of different voices are available for users to select, with options ranging from an energetic New Yorker style to a pirate style.
Avatars Gain Dancing and Sharing
Beyond functional tasks, Meta has introduced playful features for the Muse avatar, including the ability to generate videos of the character performing dances. Wang shared a demo involving a fluffy white dog avatar. In the demo, he joked about his habit of posting memes on X and how it might cause trouble with his boss.
The dog avatar suggested adding disclaimers to the memes to avoid appearing as though Meta was endorsing risky content.
To use the feature, users tap their Muse avatar, select the share button in the upper-right corner, and choose their preferred destination. This integration connects its AI assistant with its existing social media platforms. Users can create and share AI-generated content without needing third-party video editing software.
Future Hardware Integrations
Meta plans to expand the availability of its Muse AI agent beyond software applications. When this device arrives, the digital avatar will appear on a small screen to facilitate conversations and task management. Muse is intended for users aged 18 and older.