SuperwhisperSSuperwhisper
1h ago

Verbatim option for Scribe (keep filler words, false starts and repetitions)

Current behaviour: Scribe transcripts come back cleaned. Filler words ("euh"), false starts and repeated words are removed, already in the word segments, even in a mode with no AI processing. The same file had about 200 "euh" and 184 repeated words in September, and almost none since October. Ultra Cloud and S1 do the same. What I'd like: a per-mode setting, for example "Verbatim transcript (keep filler words)", that asks Scribe for a verbatim transcript (ElevenLabs' no_verbatim parameter set to false). Why it matters: I edit long podcast interviews (French, 30 to 60 min) from the word timestamps. Hesitations and repetitions are exactly what I need to see in order to cut them. For editing, subtitling or research, the raw verbatim transcript is the valuable part, and the LLM step can still clean the text for people who want it polished.
PendingPending