ideasman42 / nerd-dictation

Simple, hackable offline speech to text - using the VOSK-API.
GNU General Public License v3.0
1.21k stars 104 forks source link

Japanese models always produces spaces in the output #125

Open Madoshakalaka opened 1 month ago

Madoshakalaka commented 1 month ago

I tried both vosk-model-small-ja-0.22 and vosk-model-ja-0.22 and they always sprinkle spaces in the sentences (at word boundaries I guess). Like this:

毎日 夕方 に 買い物 に 行く (instead of 毎日夕方に買い物に行)

I wonder if there is a setting I can do. Especially Japanese don't have spaces at all, so if there is a setting that just removes whitespaces, it would help a lot.