I'm attempting to run my sdrtrunk recordings through whisper so they're searchable. I've run into an issue where recorded calls, particularly a single conversation are split over multiple mp3 files. I think the JMBE vocoder gives up a little too easily, even when audio is intelligible otherwise. This obviously causes issues with words split between two files, but also issues with the transcriber having no context for the previous conversation. Is anyone else having these troubles? What are you using to fix it?
