v2.0 , Android Available Now

Download Abserny

The complete voice-first vision assistant for Android. Gesture-driven, AI-powered, bilingual, works fully offline.

Point your camera here to download

Or open this link on your device ↗

Source Code All Versions →

Total Downloads , live · GitHub

Median Latency 1,430ms online · Gemini

Offline Latency 380ms ML Kit fallback

Spatial Term Rate 87% directions in output

Detection Modes 4 Scene · Object · Read · People

What’s Included

Everything built into Abserny v2.0, from the spoken onboarding to the offline AI fallback.

Voice-First Onboarding First launch speaks a language picker and interactive gesture tutorial. The user must perform each gesture to advance, zero sighted assistance required from the very first second.

Gemini 2.0 AI Vision Google Gemini 2.0 Flash Lite delivers spatial scene descriptions with a 1,430 ms median latency. Hazards mentioned first, spatial direction words always included, preamble-free output.

Offline ML Kit Fallback Automatic silent failover when offline. On-device image labeling and OCR via ML Kit, 380 ms median latency, 0% failure rate. No API key required for offline use.

Four Detection Modes Scene, Object, Read, and People, swipe left or right to cycle. Each mode has its own tailored AI prompts, optimized per mode and per language for the highest-quality descriptions.

Arabic & English Full bilingual support: separate Gemini prompts, native TTS voices (ar-SA / en-US), RTL layout in Arabic mode. Language chosen during onboarding and changeable via Settings at any time.

Spoken Settings Menu Triple tap opens a fully spoken settings overlay, navigate by swipe, confirm by double tap, close by long press. Replay the tutorial or change language with no visual interaction whatsoever.

Installation Steps

Step 1 , Download

Visit the GitHub Releases page and download the latest .apk file to your Android device.

Step 2 , Allow Installation

On Android, go to Settings → Security and enable "Install from unknown sources" for your browser or file manager.

Step 3 , Install the APK

Open the downloaded file and tap Install. Grant camera and microphone permissions when prompted.

Step 4 , API Key (Optional)

For online Gemini AI, add your Google Gemini API key in hooks/useDetection.js. The app works fully offline without it via ML Kit.

System Requirements

Android

Android 8.0 (Oreo) or higher. 2 GB RAM minimum, 4 GB recommended. Approximately 300 MB storage for app and model files.

iOS

Architecturally compatible. Release pending Apple Developer Program enrollment. No code changes anticipated, follow GitHub for updates.

Camera

Rear camera required for detection. Works with standard Android rear cameras, no minimum megapixel requirement.

Network

WiFi or mobile data for Gemini AI (online mode). No network required for ML Kit offline fallback, the app always works.

Release Timeline

✓

Architecture & Core Systems

Hooks architecture, FSM, gesture engine, speech queue, complete

✓

Voice-First Onboarding (v3)

Language picker and interactive gesture tutorial, complete

✓

Bilingual Support & AI Prompts

Arabic/English prompts, TTS voices, RTL layout, complete

✓

Android Public Release v2.0

Available now on GitHub Releases

○

AbserneyVision Model (80%+ accuracy)

Dataset expansion in progress

○

iOS Release

Pending Apple Developer enrollment

Stay Updated

Watch the repository for iOS updates, model improvements, and new version announcements.

Watch on GitHub All Releases