Descript Transcription

Podcasters who edit by deleting text rather than cutting waveforms.

Descript is best known as a text-based audio and video editor, with transcription as the entry door. Their in-house ASR is competitive with Whisper, and the killer move is that editing the transcript edits the underlying audio — delete a sentence in the doc, the waveform follows. Pure transcription users will find the editor's gravity hard to ignore.

Descript took a heretical position years ago — what if you edited audio by editing a transcript? — and the entire product still flows from that bet. You drop in an episode, the ASR transcribes it, and from there the document and the waveform are the same artefact. Delete a paragraph and the audio cuts to match. Re-record a sentence with Overdub and the timeline updates to match. For podcasters who think in language rather than waveforms, this is genuinely transformative. The transcription engine itself is in line with Whisper-class accuracy on clean English, with multi-language support across the major European and Asian languages. Speaker labels work reliably on clean multi-track recordings and drift on noisy single-track sources, which matches every other tool in the category. The Free tier covers a real hour per month at 720p with basic AI, Hobbyist is $24/mo ($16 annual), and Creator is $35/mo ($24 annual) with 4K export and the full AI suite. The trade-off is that if all you want is a transcript file, Descript is wildly over-tooled for the job — the export flow buries the option behind the editor, and the constant marketing of Underlord credits and AI features can feel like noise. For anyone who wants one tool to record, transcribe, and edit, it is still the obvious pick.