Synthetic data, noise augmentation, or just more recordings?
93
Synthetic data, noise augmentation, or just more recordings?
Tried wake-word training yesterday. Works, with a couple of rough edges.
Profiling first saved me a week of guessing.
@Leo_Mod Open-source repo link is in my bio, PRs welcome.
@wanjiruW Have you tried chunking audio at 20 ms frames? 👍
Doing the same with wake-word training, but with a local model, and it feels much more private.
Well said.