VoxRT

VoxRT: On-Device Voice AI SDK

VoxRT is an on-device voice AI software development kit designed to deliver secure, low-latency voice capabilities without relying on cloud infrastructure. The platform provides developers with tools for wake words, keyword spotting, and speech-to-intent, all running locally on devices to ensure maximum privacy, predictable latency, and minimal battery consumption.

Key Features

• Custom Wake Word: Train specific wake phrases robust to noise and accents with tunable confidence scoring.
• Speech-to-Intent: Maps speech directly to structured intents in a single inference without needing a transcription stage.
• Voice Activity Detection: Always-listening, power-efficient detection that gates wake-word and keyword spotting for optimized battery life.
• Custom Keyword Spotting: Detects a fixed vocabulary of voice commands with lower latency and compute requirements than full speech recognition.
• Private by Design: Audio never leaves the user's device, and models are encrypted at rest with no cloud round-trips required.
• Highly Optimized Runtime: Built from scratch in Rust, featuring production binaries in the hundreds of kilobytes and INT8 quantization for predictable latency.
• Cross-Platform Compatibility: Runs seamlessly across iOS, Android, and macOS, with support expanding to microcontrollers, automotive systems, and wearables.

Use Cases

• Mobile applications requiring secure, offline voice control and command execution
• Smart home devices and IoT hardware needing low-latency, on-device wake word detection
• Automotive and wearable technology seeking power-efficient voice interfaces without cloud dependency

Getting Started

• Website: https://voxrt.com/

VoxRT empowers developers to integrate robust, privacy-first voice features into their applications and devices, offering a highly efficient alternative to cloud-dependent voice engines while maintaining exceptional speed and accuracy.