A 13M ASR Conformer running on an ESP32-S3
wunschpunsch3D · reddit · 2026-07-21
The author shares a project that runs a 13.1M-parameter speech-recognition Conformer on an ESP32-S3 microcontroller.
Key details:
- It is a distilled and quantized version of NVIDIA's small Conformer model from Hugging Face
- The model fits into 14MB of flash, plus 256KB of SRAM and 4MB of PSRAM
- It can transcribe about 8 seconds of audio
- The first attempt took 10 minutes to transcribe 5 seconds; this version is much faster, though still slow
- Whisper Tiny was also tested, but was so much slower that it was not further optimized
- The chip's 8-bit hardware acceleration helps make ML feasible on the platform
The author notes that quantization increased word error rate by about 3% on Hugging Face ASR benchmarks, and argues that more work should go into efficiency so hobbyists can run models on affordable hardware.
More from Embodied
- FocusVLA routes robot camera views by predicted interaction intent across task stages — iscas-cloud · 2026-10-26
- Pandora M1: first prototype of a physical device for distributed GPU monitoring — tarantulae · 2026-10-26
- $150 modular Linux handheld Mecha Comet opens preorders — pocket SSH and Claude Code included — Paimaamu · 2026-10-26
- Apple Unveils All-New iPad mini: OLED Display, A20 Pro, Biggest Update Ever — CurieuxExplorer · 2026-10-26
- Purdue's GPS-free autonomous drone tracks runners through dense forests using on-chip LiDAR AI — KordingLab · 2026-10-26
- Apple lists OLED iPad Mini for sale on its online store — etothet · 2026-10-26