lokutor-ai/oido-v2-realtime-int4
Automatic Speech Recognition • Updated • 63
Oído: speech recognition on a $5 microcontroller
Open-vocabulary English and Spanish speech recognition for the ESP32-S3: v2 runs faster than speech on one core. Code: github.com/lokutor-ai/oido
Note Real-time v2 (A: best accuracy, B: fastest), int4
Oído v2 demo
🎙Speech recognition on a $5 chip, in your browser
Note Try it in your browser: the chip's engine in WebAssembly
lokutor-ai/oido-ctc-small-stream-int8
Automatic Speech Recognition • UpdatedNote v1 streaming model (int8)
lokutor-ai/oido-ctc-small-int8
Automatic Speech Recognition • Updated • 6Note v1 most accurate English model (int8, utterance mode)
lokutor-ai/oido-ctc-small-int4
Automatic Speech Recognition • Updated • 3Note v1 English, smallest (int4, utterance mode)
lokutor-ai/oido-es-ctc-small-int8
Automatic Speech Recognition • UpdatedNote Spanish (streaming + full context)
Oído: Open-Vocabulary Speech Recognition on a $5 Microcontroller without a Neural Accelerator
Paper • 2610.13447 • PublishedNote The paper (v2 describes the real-time models)