EVA
EVA OS
hero-bg

Multimodal AIOS for Next-Generation Hardware

Create a real brain with “life” and “high intelligence” for your hardware

EVA OS
Intelligent Multimodal Core
Our Advantages
EVA OS is AutoArk’s core multimodal system, combining speech, video, and vision with ASR/TTS and system orchestration to enable real-time, natural interaction. We welcome developers to join as Maintainers.
Multimodal Interaction
Voice, vision, and text work seamlessly together for natural real-time interaction.
视频
EVA OS
音频
Speaking
EVA OS
文本
你是一名心理疏导机器人 |
EVA OS
Low-Latency Speech & Visual
ASR, translation, and TTS deliver fast, stable, and interruption-free communication.
LatencyLatency
Custom Voices & Emotion
Create unique voice styles with fine-grained emotional expression.
TTS
Zero-Code Multimodal Agents
Build powerful multimodal agents effortlessly with simple drag-and-drop tools.
Zero Code
EVA OS
Full Hardware Development Support
Complete open-source firmware for all
supported boards.
ESP32、RK 3562/3562/3568...
Product
EVA OS
Full Hardware Development Solutions
Hardware Product Use Cases
Aqi Companion Robot
Aqi Companion Robot
Powered by EVA OS with full-stack multimodal capabilities, delivering sub-350ms real-time voice and vision interaction.
Aqi Companion Robot
Kidodo Early Education Robot
Kidodo Early Education Robot
Designed for children aged 0–10, combining emotional companionship and science-based learning with AI-driven interactive stories.
Kidodo Early Education Robot
Confidential Smartphone Assistant Project
Confidential Smartphone Assistant Project
A next-generation AI assistant built on EVA’s multimodal model, enabling natural dialogue, real-time perception, and intelligent decision-making. Currently in confidential development.
Confidential Smartphone Assistant Project
Confidential AR Glasses Project
Confidential AR Glasses Project
Lightweight AR intelligent eyewear with visual understanding, spatial sensing, and multimodal interaction. Designed for seamless real-world integration. Technical details remain confidential.
Confidential AR Glasses Project
Confidential In-Vehicle Assistant Project
Confidential In-Vehicle Assistant Project
An intelligent in-car assistant combining voice and visual capabilities for real-time perception and multi-scene assistance. Developed in partnership with automotive teams.
Confidential In-Vehicle Assistant Project
EVA OS
Development History
EVA OS Timeline

01 Aug 2024

Multimodal Base Model Released
Launched our foundation multimodal model built on Transformer architecture, supporting long-context generation up to 2M tokens.

05 Sep 2024

Aqi Robot Debuts at Overseas Expo
Aqi showcased ArkModel 2.0 with GPT-4o-level latency, multimodal perception, and interactive capabilities, attracting significant attention.
Aqi Robot Debuts at Overseas Expo
Aqi Robot Debuts at Overseas Expo

16 Sep 2025

Kidodo Pre-orders Exceed 10,000 Units
“Kidodo AI Story Companion” surpassed 10,000 pre-orders and joined the LUMI Global Education Initiative.
Kidodo Pre-orders Exceed 10,000 Units

08 Dec 2025

EVA EVA OS Developer Release
Released desktop and mobile demos, streaming solution configs, video/audio pipelines, TTS voices, Agent tools, MCP components, open-source docs, and GitHub access for developers.

Upcoming

Upcoming
New features and capabilities coming soon. Stay tuned.
🔌 Embedded SDK: ESP32/RK/MCU hardware-level deep adaptation
☁️ Cloud-Edge Collaboration: Cloud-side compute scheduling + device-side precise instruction dispatch
🧠 Intelligent Memory System: Device-level short-term interaction memory + long-term preference memory, multimodal technology
🔧 MCP Tool Extension: Open third-party tool access interface
🎭 Custom Digital Avatar: Support photo generation of custom avatar
🏢 Enterprise-level Capabilities: Multi-model hybrid deployment + high-concurrency capability
An open-source, real-time multimodal agent engine that lets your devices hear, see, and act.
Visit GitHub