akdeb

ElatoAI: Build Realtime Voice AI Devices with ESP32

ElatoAI connects ESP32 devices to realtime speech-to-speech AI through secure WebSockets and edge servers. It suits makers building conversational toys or devices who can work with Arduino firmware, cloud services, and external AI model providers.

ActiveTypeScriptAIReal TimeAI Agents
View on GitHub Compare
Stars
2k
Forks
256
Stars, last 30 days
+61
Last commit
12 days ago
Contributors
4

Our take

Activeregular commits

Health score

23/100

How it is scored
Activity
35
Community
16
Issues
21
Pull requests
20

+61 stars in the last 30 days

ElatoAI is an actively updated, specialized starting point for ESP32 voice-device projects, but its development and maintenance are highly concentrated in one contributor and it has had no release in the past year.

Good fit if

  • You are building an ESP32-based voice device and want integrations with hosted realtime speech models.
  • You can work across Arduino firmware, edge/server code, and services such as Supabase and external AI providers.
  • You value a project with regular recent commits and are comfortable evaluating changes from a small contributor group.

Look elsewhere if

  • You need a released, versioned package or evidence of broad contributor ownership before adopting it.
  • Your device must detect speech interruptions locally, a limitation the README calls out.
  • You need a complete local-only speech stack; the described architecture relies on edge servers and provider APIs.
All health signals
Last commit2026-09-25 (12 days ago)
Commits, last 90 days12
Releases, last 12 months0
Contributors4 (top contributor: 98% of commits)
Issues closed, last 90 days0
Pull requests merged, last 90 days0
Project age18 months

Checked on 2026-10-08 with the GitHub API.

Project history

Daily GitHub snapshots since Oct 7, 2026. History covers stars and forks; health signals are not tracked historically.

Selected snapshot

Stars

2,118

Forks

263

Change since Nov 14, 2026: 0 stars·0 forks

Overview

ElatoAI is a hardware-and-software project for building voice-enabled devices around an Arduino-compatible ESP32. The device streams audio over secure WebSockets to an edge server, which connects it to realtime voice models or assembles a speech pipeline from speech recognition, language, and text-to-speech services.

It is aimed at developers who want to make conversational toys, companions, or other connected devices without running speech models on the microcontroller. A Next.js web app and Supabase support agent and device management, while server options include Deno Edge Functions and Cloudflare Workers.

Key Features

  • Connects ESP32 devices to realtime voice services including OpenAI Realtime, Gemini Live, Grok Voice Agent, ElevenLabs, Hume, and Boson.
  • Offers Cloudflare Workers and FastAPI options for composing speech pipelines from multiple model providers.
  • Streams audio over secure WebSockets and uses Opus compression.
  • Includes device setup features such as Wi-Fi configuration, touch or button controls, and over-the-air updates.
  • Provides a web app for managing devices and AI characters, with conversation history and transcripts stored through Supabase.
  • Supports tool calls from the device through edge functions.

Use Cases

  • Makers building interactive toys or companions: Connect an ESP32-based device to hosted voice models without requiring the device to run those models locally.
  • Prototyping voice interfaces on embedded hardware: Try different speech-to-speech providers or assemble a pipeline using the server options.
  • Teams managing multiple connected devices: Use the web app and the documented Cloudflare approach for device and agent management.
  • Developers exploring conversational hardware: Use the Arduino and PlatformIO paths when a project needs physical controls, speakers, and microphone-based interaction.

What you need

Detected in the repository

  • TypeScript (main language on GitHub)

Getting Started

Start with the Quick Start guide, then choose the PlatformIO or Arduino IDE setup. See the README for architecture and server options.

License in Plain Words

No standard license was detected. Without a license, the default copyright rules apply: you may read the code, but not reuse it. Check the repository or ask the authors.

Alternatives

  • xiaozhi-esp32-server: Both serve AI voice interactions on ESP32 devices, but xiaozhi-esp32-server is a self-hostable backend for xiaozhi hardware with device management features.
ProjectLanguageLicenseStarsStatus
ElatoAITypeScript–2kActive
xiaozhi-esp32-serverJavaScriptMIT10.7kActive

Considerations

  • This is a multi-part system, not just an ESP32 firmware library. A deployment may involve a Next.js frontend, Deno or Cloudflare edge functions, Supabase, and credentials for an AI provider.
  • The README reports a 2-3 second cold-start connection delay, says conversations have been tested up to 17 minutes, and notes that edge execution can stop when wall-clock limits are reached.
  • Speech interruption detection on the ESP32 is not supported, according to the README.
  • Repository health data shows regular recent commits, but no releases in the past year. Most contribution activity comes from one person, and no issues were closed in the measured recent period.
  • The README says the project is MIT-licensed, while the supplied GitHub license metadata is marked NOASSERTION. Check the repository's license file and confirm the applicable terms before relying on that claim.

Found this useful?

Share it with someone who would like ElatoAI.

Comparisons