AI Briefing
KO

Ultra-small TTS 'sanoTTS' Released

·2026.09.04 07:01

Key point

The ultra-small TTS model sanoTTS with 294k parameters has been released, enabling real-time speech generation even on $3 chips.

Details

Developer Ampixa has released sanoTTS, an ultra-small, complete TTS stack with 294k parameters. The model is 337KB in size and is designed to run on $3 microcontrollers (ESP32) without an NPU.

Performance and Features

  • Parameter Size: Ranges from 294k to 2.2m, which is 1000x smaller than Kokoro and 9000x smaller than Voxtral TTS.
  • Voice Quality: The 1.5m model achieves a WER of 2% and UTMOS of 4.10, attaining a higher SCOREQ score (4.13) than Inflect Nano (4.63m) and Kitten TTS (15m).
  • Real-time Processing: On an ESP32 microcontroller, it achieves an RTF (Real-Time Factor) of 0.225, generating 4 seconds of audio per second.
  • Multilingual Support: Supports 11 voices and 6 languages, and can run on the web via WebAssembly.

Deployment and Usage

Released as open source on GitHub and Hugging Face, it can be easily integrated into web projects via npm install sanoTTS. It presents new possibilities for implementing voice interfaces on low-power devices or edge devices.

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.