Infatoshi
Featured
netherite
A from-scratch C/CUDA reimplementation of Minecraft 1.11.2 that is bit-verified against the original Java game and includes batched CUDA reinforcement learning support.
AI repository intelligence
FR-AI / ANALYSISWhy it is worth attention
It achieves a complete voice-to-text pipeline on constrained hardware using static memory (no malloc/free), a single-threaded event loop, and Linux framebuffer rendering — all in ~4,800 lines of C, making it a reference design for embedded AI assistants.
Who it is for
- Embedded Linux developers building voice‑controlled devices
- Hobbyists experimenting with AI on low‑cost ARM boards
- Educators teaching embedded systems and real‑time audio processing
- Product designers prototyping kids' or elderly‑friendly voice interfaces
Use cases
- Kids' Q&A: press button, ask a question, see the answer on screen
- Kitchen helper: voice‑query recipes while hands are busy
- Elderly companion: voice‑first interaction for seniors uncomfortable with smartphones
- Desk assistant: quick voice queries for exchange rates, weather, or reminders
Strengths
- 100% static memory allocation eliminates runtime fragmentation on limited RAM (~256–512 MB)
- Single‑threaded 50 ms event loop simplifies concurrency and reduces context‑switch overhead
- Uses Linux Framebuffer directly instead of a heavy GUI toolkit, supporting both 32bpp and 16bpp pixel formats
- Comprehensive error handling with a 5‑state machine that gracefully recovers from audio, WiFi, HTTP, timeout, or server failures
Considerations
- Requires the specific Rockchip RK3506 development board and its BSP (Buildroot/Yocto) — not portable to other SoCs without rework
- Single‑utterance recording capped at 60 seconds and no voice activity detection (VAD) yet (on roadmap)
- Dependent on an external AI server via HTTP — voice recognition quality and availability rely on a remote service, and the README uses a placeholder IP address
README quick start
Quick Start
Description
RK3506 Voice Robot An embedded AI voice assistant running on the Rockchip RK3506 development board. Pure C, single-threaded event loop, zero dynamic memory allocation.
Related repositories
Similar projects matched by category, topics, and programming language.
nsdkinx
Featured
twell
Twell is a tiny, embeddable, zero-allocation physics engine for interruptible, Apple-style UI animations, available as a single-header C library with Python bindings.
sqliteai
Featured
waste
WASTE is a C-based inference engine that streams MoE experts from disk to run a 2.78 trillion-parameter model on a consumer laptop with 64 GB RAM.