Well, another year has nearly passed and I still have not solution for voice controlling my home automation system / Node Red. Hell I spend so much time for trying out solutions and non of them worked for me. Maybe you know something I don t or have any working solutions ?
Here is what I was searching for:
Must haves:
- German STT
- Own hot word
- No monthly fees to spend
- Filter for background noises
Should haves:
- API to get sound files / streams form decentral devices like PI Zero W or mobil phone
- Docker support
- Easy installation
Nice to haves:
- Local recognition (non cloud)
I tried and gave up:
Raspberry Pi 4 -> Coqui v0.9 (best so far):
- no hot word support
- bad and old language model
- fails on small background noises
- node red plugins / implementations outdated
Raspberry Pi 4 -> Voice2Json:
- only self trained phrases / no full STT
- some technical problems I could not solve
- node red plugins / implementations outdated
- project seems to get no more updates
WSL2 / Docker -> Rhasspy:
- only self trained phrases / no full STT
- could not get it to work with Docker (sound recording)
- phrases definition outside node red