ScribeDrop

ScribeDrop turns a recording into a written transcript, and works out who is speaking. Drop in a file of a meeting, an interview or a conversation and you get the words with proper punctuation, split into Speaker 1, Speaker 2 and so on. Type a real name over any speaker and it changes everywhere. Everything runs on your Mac. Nothing is uploaded. No account, no subscription. What you get Who said what. Tell it how many people are talking, up to ten, and it splits the conversation by voice. It is more accurate if you give it the number. Punctuated text you can read, not one long block. Click any time to hear that spot. A play bar lets you scrub through the recording, and the transcript follows along. Fix the speaker changes. If it gave a few words to the wrong person, click the word and say "from here on, this is Speaker 3" or "up to here, this is Speaker 2". The words move across and the boundary lands exactly where you put it. Missed someone? Add a speaker and hand them their words. Fix spelling by editing the words in any block. Save to edit later. Your transcript, names, edits and times are kept in one file you can reopen whenever you like, so you can come back when you know who everyone is. Export as text, Markdown or PDF, with or without times. Or copy the lot. Handles long recordings. Your Mac stays awake until it is done, and you can stop at any time. Good to know The first time you use it, ScribeDrop downloads its speech and voice models, roughly 1 GB, and keeps them on your Mac. Long recordings take time. A recording of an hour or more can take a good while to work through, depending on your Mac. It is not perfect, and it doesn't pretend to be. Speech recognition can mishear words, and voice separation can mix people up, most often when two voices sound alike (two men of a similar pitch, say) or when people talk over each other. In testing it got most of a five person conversation right and slipped in places between two similar voices. That is what the listening and fixing tools are for. Listen back and check anything important before you rely on it. Don't use it as the only record for legal, medical or other critical purposes without checking it. Works best with English. You can tell it the recording is in English or let it work out the language. You'll need a Mac with Apple Silicon (M1 or newer) running macOS 14 or later. It isn't a signed, notarised app from the Mac App Store. The first time you open it, macOS may ask you to confirm you trust it. Right click the app, choose Open, then click Open. You only do that once. Credits ScribeDrop uses OpenAI Whisper and Argmax WhisperKit and SpeakerKit (MIT licences), and speaker models based on Pyannote, released under CC BY 4.0. Full licence texts come in the zip.