sqlpxc/airhand
REMOTE DESKTOP OVER QUIC WITH GPU
Airhand is a high-performance remote desktop with QUIC transport. It uses GPU for AV1/HEVC/H.264 encoding up to 120 fps and ensures low latency by skipping frames during network congestion.
// KEY FEATURES
- Hardware AV1/HEVC/H.264 encoding on GPU (NVENC, Quick Sync, AMF, VideoToolbox, MediaCodec)
- QUIC transport with BBR and direct NAT traversal with relay fallback
- Local cursor rendering and frame skipping during network congestion
- Web console in browser and Windows system service for login screen and UAC
// DEEP DIVE: HOW IT WORKS
Airhand consists of a host and controller connected via a self-contained relay server deployed with one Docker Compose command. The screen is encoded on the host GPU (AV1/HEVC/H.264, low-latency, no B-frames), sent over QUIC with BBR, and decoded on the client GPU (WebCodecs/MediaCodec). The host limits bytes in flight and jumps to the latest frame on congestion; the cursor is rendered locally.
Key differences: hardware acceleration on both ends, automatic codec selection, direct NAT path with seamless fallback. Protocol v9+ supports version negotiation. Controller available as native app or web (Docker), console with tabs and multi-screen overview.
Status 0.2.3 (WIP), protocols <v9 incompatible. Requires GPU with NVENC/Quick Sync/AMF/VideoToolbox/MediaCodec. Windows service for login screen/UAC. Relay self-hosted. Supported: Windows, macOS, Linux, Android; controller also in browser.
// USE CASES
Best for: For engineers and admins needing remote access with minimal latency on any devices
- Control remote PC from login screen and UAC via Windows system service
- Connect to machine from browser without client install via web console in Docker
- Monitor multiple computers simultaneously in tabs with multi-screen overview
New repositories every 30 minutes
REDDYX AI scans GitHub 24/7 and ships the best AI/ML/Web3 projects to Telegram.
Join on Telegram