---
title: KEEP Listen
colorFrom: indigo
colorTo: green
sdk: gradio
app_file: app.py
pinned: false
license: other
short_description: Recognizes 10 spoken English words as it listens.
---

# KEEP Listen

AMEFORGE demo: a small model (about 265,000 parameters) recognizes **10 spoken English words**
(yes, no, up, down, left, right, on, off, stop, go) and shows how its confidence rises while listening.

- Measured accuracy: about 97% on never-heard voices (Google Speech Commands v0.02, official test lists).
- Limits: only these 10 words; any other sound is classified as one of them. Recognizing is not understanding.

Proprietary AMEFORGE model. Model weights and model code are not public.