IrregularChat Q&A

Anyone have good tips or pipelines for Speech to text training? I’ve been working with whisper models and LoRA-whisper. I am trying to improve my transcriber for public safety radio traffic. Even whisper large-v3 struggles with the vernacular / jargon and phrasing. It has a lot of repeatable or canned phrasing though. So far my training has yet to beat baseline though