Transcripts and Ask
This section covers transcripts, summaries, and Ask. Text from the publisher needs nothing installed. Text the app generates needs a model download, and the file is checked against a published hash before those screens are available.
Nothing in this section is uploaded. Ask can still be wrong. Read the source cards before you trust an answer.
TranscriptsPublisher text and local speech-to-text, kept apart.
ModelsSpeech model, then Quality or Fast. Optional.
StudioOne persistent queue for transcripts and summaries.
SummariesA recap from a transcript, checked against that transcript.
AskA question, source cards, and no memory after you quit.
ExportOne episode as Markdown, including the timed transcript.
Playback, bookmarks, library search of titles, voice, and focus sounds do not need a model. Search inside spoken words needs a transcript, which you may already have from the publisher.
