What is it?

A compact neural text-to-speech module that runs entirely offline on a microcontroller. It converts any text into natural, human-like speech without a cloud connection — just power it and play.

Built on the Himax HX6538 (Arm Cortex-M55 + Ethos-U55), it runs a neural-network TTS engine directly on-chip, delivering real-time speech generation in a tiny, battery-friendly form.

Type any text through a connected interface (like CrowPanel Advance 5.0) and instantly hear your project talk back — no Wi-Fi, no latency, no privacy issues.

Why did you make it?

We wanted to make realistic voice output accessible for embedded projects — robots, IoT devices, and smart panels — where cloud access isn’t possible or secure. TinyTTS gives makers and developers a way to add speech directly on-device, simply and reliably.

What makes it special?

It’s powered by edge AI hardware (Himax HX6538 with Cortex-M55 + Ethos-U55 NPU) and runs a compressed, quantized neural network for fast, smooth text-to-speech. No internet, no delay — just pure local AI.

Key Features

• Fully offline neural TTS engine for MCU-class devices
• Powered by Himax HX6538 (Cortex-M55 + Ethos-U55)
• 3.5 mm audio output jack — connect any small speaker
• Runs efficiently on battery-powered devices
• Works with Arduino, ESP32, and CrowPanel Advance
• Generates natural speech from any text length
• Low power, compact, plug-and-play design

Use Cases

• IoT devices that speak their status
• DIY robots that narrate actions
• Smart appliances with voice alerts
• Educational and accessibility projects
• Industrial systems with spoken warnings

Package Includes

• 1 × GRC TinyTTS Hx Module
• Documentation & example code (via GitHub)

Resources

• GitHub Repository — examples & code
• Hackaday Project Page
• Hackster Project Page — demo & use cases

Ready to make your project to speak? Add the GRC TinyTTS Hx Module to your cart, connect it to CrowPanel or Arduino, and bring neural voice AI to the edge.