Seems to use the VOSK voice recognition model under the hood for most languages, but doesn't really say that upfront or how it fairs against state of the art. Vosk has significantly higher word error rate over whisper models in english but since it's an editor of sorts it probably doesn't matter that much. They are working on adding whisper as an option.
https://scribe.rip/version-1/analysis-of-automatic-speech-re...