Oído: open-vocabulary speech recognition that fits in a $5 ESP32-S3. 3.7% LibriSpeech WER, no cloud, no NPU.
85stars4forksC
asrconformeredge-aiembeddedesp-idfesp32esp32-s3microcontrolleron-device-aispeech-recognitionspeech-to-texttinyml
Real data pulled from GitHub this week. The author's original repo lives upstream.
View on GitHub