whereismyai
Whistle Esp32
Whistle Esp32 is an embedded speech recognition system running fully on an ESP32 microcontroller without network or cloud.
Whistle Esp32 runs speech recognition entirely on an ESP32-S3 microcontroller using an independent C99 implementation of the Whistle model. It listens via an attached INMP441 microphone and controls an onboard WS2812 RGB LED based on spoken commands. Users hold a button, speak a colour command in one of seven languages, and the device processes the audio locally without any cloud connection or phone. The system uses INT8 quantization on the S3 processor to maintain accuracy while keeping processing times practical. It includes tools for checking microphone input, evaluating languages, and flashing the stripped model to the board's flash memory.
Whistle Esp32 at a glance
Key features
- Runs fully offline on an ESP32-S3 microcontroller
- Recognises speech in seven different languages
- Performs automatic language detection
- Controls an RGB LED via spoken color commands
- Processes audio using int8 quantization
- Measures ambient noise floor to avoid false triggers
- Includes standalone testing tools for host computers
Use it for
- Controlling hardware lights with spoken color commands
- Running offline speech recognition on low-cost microcontrollers
- Testing embedded machine learning models on edge hardware
Alternatives to Whistle Esp32
- Esp32 Pie ASR — Faster than real time streaming speech recognition on a microcontroller alone
- Esp32 AI — Runs a language model directly on a microcontroller chip without a server.
- whisper.cpp — It provides high-performance, offline speech recognition on diverse hardware.
- Babytalk — Runs entirely on-device with no cloud or internet required.
- ESP-Claw — Runs agent runtimes directly on device chips for local decision making.
- ESP-Mosaico — Integrates an on-device AI agent runtime with an open desktop development workspace.
- Soup — Fine-tune and align models like Llama-3.1-8B on a 4 GB laptop GPU.
- Veecle — Compile and simulate firmware on actual target chips without physical hardware.
- AirLLM — Runs massive 70B+ parameter models on small 4GB consumer GPUs.
- Pulsar — It is the fastest inference engine for Qwen models on Apple Silicon.
- ODS — Automates complex local AI infrastructure setups with one command.
- LLamaSharp — Runs large language models efficiently on local hardware using C#.
All Whistle Esp32 alternatives