The Pebble app doesn’t currently give you the option to add alternative models, but most people already have models that work well for identifying speech built right into their phone via the phone’s voice dictation API, often through Google speech recognition services, or in my case, an open source alternative from FUTO.
It would be nice if users could simply set whatever default app/service is used for their system’s voice recognition as the system that the Pebble app would use for any voice transcription from the watch. Could also possibly fix the issue of local voice dictation not working when the phone is locked, given that service is often something used by voice assistants, which can (depending on the implementation) be available from the lock screen.