Post #2142290
2026-01-06 18:30 UTC
@M Great questions. The key will be how I can best utilize OpenWakeWord. That's the piece using our existing default model, will need to grow.
The challenge and a doable one, will be folks being able to train their own. So it's on my radar and also how we best get past English models as well.
All are considerations I'm committed to sorting out over time.
Replies (2)
-
@M@nerdculture.de 2026-01-06 19:12
@matthartley Thanks.
-
@zeborah@mastodon.nz 2026-01-07 00:10
@matthartley @M Train-your-own would be excellent! In the meantime, what accents has it been trained to recognise -- and/or tested on? (That might be useful info in the readme too.) I remember pre-Dragon I could get MacOS's built-in system to work sometimes if I faked the accent just right...