Show HN: Live speech-to-text on an ESP32-P4 microcontroller, no cloud

Automated excerpt

The weights are on Hugging Face: LibriSpeech test-clean 4.34 % WER, test-other 10.26 % (beam search, 4 hypotheses) 0.70× real time: each 1.28 s of audio in 0.90 s, both cores at 400 MHz a word is final about 2 s after it is spoken (median 1.9 to 2.3 s, 95 % within 4.1 s) 26.65 MB, read from a microSD card at boot in 2.9 s voice-control demos. From the end of an order to its action: median 1.84 s on the vacuum's page, 1.97 s on the factory's, at most 2.57 s.

Selected automatically from source text; not independently written or fact-checked. Read the original for full context.