2026-09-30 · America/Los_Angeles · 社区动态 · #13

Oído: speech recognition that beats Whisper-tiny, running on a $5 microcontroller (open source)

I'm part of the Lokutor team that built this. Model: NVIDIA Conformer-CTC Small (13M params, int8). It runs on an ESP32-S3 with 8 MB PSRAM, no GPU or NPU. LibriSpeech WER is 3.7 / 8.2, versus 6.3 / 15.9 for Whisper tiny.en on a laptop. Under real noise (DEMAND: car, kitchen, cafeteria) plus babble and reverb, mean WER is 8.4 vs 12.1 for Whisper tiny.en. You can try the exact chip arithmetic on your laptop mic with live_demo.py. https://github.com/lokutor-ai/oido

热度 46.9 / 100；排名与评分保留该期记录。