A neural text-to-speech model running entirely on an ESP32-S3 with Arduino-ESP32: no PSRAM, no cloud.

Type text, choose a style, and it starts speaking in about a third of a second. You can also slide controls for speed, pitch, loudness, pauses and breathiness.

• Uses about 122 KB of RAM and about 5 MB of flash
• Two voices, one female and one male; English only
• Free for non-commercial projects

Code: https://github.com/WhissleAI/steer-tts (esp32/ folder)

Model: https://huggingface.co/WhissleAI/steer-tts-nano