Jasper Model Used for the quickstart

Hi, so I’m wondering if the speechtotext_english_jasper:deployable_v1.2 model is the same model used in the quickstart guide, and if so what language model is used with it, because I’m trying to build my own jmir with custom vad configs, but the results I’m getting are not the same as the results on the quick start script for live streaming. Am I doing something wrong or are the models not the same?

Hi @HansieB,

Yes, it should be same (speechtotext_english_jasper:deployable_v1.2 model).
Could you please share the steps/script and logs so we can help better?


Jarvis_Steps.txt (1.3 KB)

Hi @SunilJB

I uploaded some bootstrapped steps I follow to build a model. Thanks for the help and any suggestions would help, (maybe I need to set the beamwidth etc…?).

ps.There no problem with the building or starting of the server the wer is just like 8-10% worse when building my own