Project Astra by Google logo

Project Astra by Google

Paid
5.0
Inputs: video, audio, textOutputs: audio, text
Type
Saas
Company
Google DeepMind

About Project Astra by Google

Project Astra is a research prototype from Google DeepMind exploring breakthrough capabilities for Google products, with the goal of building a universal AI assistant. It integrates natural interaction, proactive responses, context-aware dialogue, action intelligence, agent highlighting, tool use (Search, Gmail, Calendar, Maps), intelligent personalization, and multimodal memory. The system works on Android phones and prototype glasses, enabling cross-device conversations. A version is being developed in collaboration with the blind and low-vision community, featuring a Visual Interpreter prototype that understands objects and unfamiliar spaces, using Google products like Maps, Photos, and Lens. Project Astra powers some features in Gemini Live, such as screen sharing and video understanding.

Key Features

Natural interaction with improved audio input/output and smooth cross-language communication
Proactive responses that start conversations without interruption or time lag
Context-aware dialogue that ignores background conversation and irrelevant speech
Action intelligence to react to user needs and take actions on behalf of the user
Agent highlighting that understands objects in context with on-screen highlights
Tool use integration with Search, Gmail, Calendar, Maps, and interface control
Intelligent personalization with personalized reasoning and memory of user preferences
Multimodal memory that integrates different data types and remembers past interactions
Content retrieval from PDFs, recipes, etc. to provide tailored help
Cross-device memory allowing conversations to continue across Android phones and prototype glasses

Pros & Cons

Pros
  • Generates content significantly faster than the fastest previous model
  • Proactive and context-aware without interrupting or lagging
  • Supports natural, smooth communication across languages
  • Can ignore irrelevant background noise and speech
  • Learns and retains user preferences for personalized responses
  • Works across devices with memory continuity
  • Designed with input from the blind and low-vision community for accessibility
Cons
  • Currently a research prototype with limited availability (waitlist)
  • Pricing model is not publicly disclosed (contact required)
  • Requires compatible hardware (Android phone or prototype glasses)
  • Some features may still be in development (e.g., glasses form factor)

Best For

Assisting blind and low-vision users by describing surroundings and identifying objectsScreen sharing and video understanding for interactive assistance on AndroidPersonalized shopping recommendations based on user preferencesRetrieving content like PDF user manuals or recipes to provide tailored helpUsing Google Maps, Photos, and Lens to explore unfamiliar places

Alternatives to Project Astra by Google

FAQ

What is Project Astra?
Project Astra is a research prototype from Google DeepMind exploring breakthrough capabilities for Google products, aiming to build a universal AI assistant. It includes features like natural interaction, proactive responses, context-aware dialogue, and multimodal memory.
How can I try Project Astra?
You can join a waitlist on the Project Astra website. There is also a Trusted Tester program for blind and low-vision users to access the Visual Interpreter prototype with live human oversight.
What devices does Project Astra work on?
Project Astra works on Android phones and prototype glasses. It supports cross-device memory so you can switch devices and continue the same conversation.
What capabilities does Project Astra have?
Capabilities include natural interaction, proactive responses, context-aware dialogue, action intelligence, agent highlighting, tool use (Search, Gmail, Calendar, Maps), intelligent personalization, personalized reasoning, content retrieval, and multimodal memory.
Is Project Astra available for blind and low-vision users?
Yes, a version called Visual Interpreter is being developed in collaboration with the blind and low-vision community. It understands objects and unfamiliar spaces, using Google Maps, Photos, and Lens. A Trusted Tester program is available for early access.