# MicroWakeWord Trainer (Nvidia)

MicroWakeWord Trainer (Nvidia) is a self-hosted ai app. A Docker-based, NVIDIA-GPU-accelerated web app for recording audio and training custom wake-word models for MicroWakeWord, openWakeWord, and NanoWakeWord, so you can create and compare wake words for Home Assistant... It is open source and self-hosted, installable on Unraid.

Canonical page: https://ourlocalhost.com/app/3059/microwakeword-trainer-nvidia

## Description

A Docker-based, NVIDIA-GPU-accelerated web app for recording audio and training custom wake-word models for MicroWakeWord, openWakeWord, and NanoWakeWord, so you can create and compare wake words for Home Assistant Assist or the Tater assistant.

## Platform availability

Available on 1 of 10 platforms: Unraid.

## Health score

71/100 (Good), based on 4 of 5 factors.
- Maintained: 100 (last commit 14 days ago)
- Popular: 45 (169 GitHub stars)
- Easy to install: 35 (1 install platform, install notes, Docker image)
- Security check: 82 (no serious findings, 3 minor, CI, tests)

## Links

- Unraid: https://ca.unraid.net/apps/microwakeword-trainer-nvidia-1ofnkd60c5avkm?q=microWakeWord_Trainer-Nvidia
- Official website: https://taterassistant.com

## Common questions

### Does the Docker trainer support NVIDIA/CUDA acceleration?

Yes, the project trains custom microWakeWord models in Docker with NVIDIA/CUDA acceleration.

### Can the trainer use samples captured from a device?

It includes device-captured wake audio as real samples, along with TTS ensembles and reviewed false-wake negatives.

### Does the interface show training progress?

Yes, it provides live training logs.

### Can trained wake words be linked to Tater Native satellites?

It provides local wake-word links for Tater Native satellites.

---
Source: Our Local Host (https://ourlocalhost.com/app/3059/microwakeword-trainer-nvidia), the self-hosted and open-source app directory.
