So this is, uh, Makas is a… hang on, let me start that again.
Retake
Makas edits a podcast like a document.
Makas writes down every word, right on your Mac. AI catches misheard names and suggests what to cut, with a reason for each. You read it through, keep what you want, and Makas renders the rest.
Cutting 14 suggested · 37 pauses · 2 manual
-
1
Transcribe
Runs on this Mac: writes down the words, tells the speakers apart, and finds long pauses. Uses the names & terms below.
Kubernetes, Postgres Re-transcribe English 2 speakersName the speakers. Names show in the transcript and help the AI.
MuratHamza -
2
Fix transcript
AI suggests fixes for misheard words and names. Accept or reject them in the transcript.
Suggest corrections Accept all (1) -
3
Find cuts
AI suggests off-topic, dull, mistake and retake sections.
keep the jokes Find cuts with AI -
4
Render
Trims the video using every active cut.
Render video…
MuratSo what finally made you switch?
HamzaHonestly? The bill. We were paying for three
Cooper NettiesKubernetes“Kubernetes” fits the talk about clusters.
clusters to run one app.
MuratThree clusters. For one app.
HamzaStaging, production, and one nobody remembers making.
Hamza restarts the sentence right away, and the second take is cleaner.
So we moved it all onto one, uh, onto a…
So we moved everything onto one box running
post grassPostgresThe database, misheard as two words.Accepted.
A football tangent, unrelated to the episode.
MuratDid you see the match last night, by the way?
HamzaDon't. I'm still not over that penalty.
MuratHe didn't even touch him!
Anyway. Wasn't one box scary?
Reading out the old price tiers one by one drags on.
HamzaA little. But look at the old bill: the first tier was forty dollars a month, the second was a hundred and ten, and then egress on top of that…
…which came to another twenty or thirty, depending on the month.
It's been eight months now. It hasn't gone down once.
A sample episode after Find cuts. Switch a cut with Keep instead or Cut it, or click a wavy underline to review a suggested fix. The Edited time follows along.
How it works
Four buttons on one side of the editor, the transcript on the other. Every suggestion lands in the transcript first, and you have the last word on all of it.
-
1
Transcribe
Makas writes down every word, right on your Mac, and tells the speakers apart so you can name them. Long pauses are found and trimmed to a natural beat. Add the names and terms you use, and it spells them right.
-
2
Fix the transcript
AI reads the transcript for words that were misheard, like names and jargon, and suggests fixes:
Cooper NettiesKubernetes. Accept or reject each one, or double-click any sentence and fix it yourself. -
3
Find cuts
AI marks the parts worth cutting and says why: off-topic tangents, dull stretches, mistakes, retakes and pre/post-show chatter. Tell it what you want, like “keep the jokes” or “aim for about 45 minutes”.
-
4
Review
Keep or cut each suggestion, shift-click sentences to cut your own, nudge a cut's edge by a single frame, and play the edit with the cuts skipped. Undo covers everything, a whole round of suggestions included.
-
5
Render
Makas joins what's left with short audio fades, so no cut clicks. Export a video ready for YouTube, a smaller file to share, or just the audio for your podcast feed.
Cuts land between words, never on them
The AI only chooses which sentences to cut. Makas places each cut in the middle of the pause between two sentences, so no word ever gets clipped.
Fixes are just as careful: a suggested fix has to quote the words it replaces exactly, or it's dropped. Every word keeps the timing it was spoken with.
HamzaSo we moved everything onto one box running Postgres.
MuratDid you see the match last night, by the way?
Off-topic
HamzaDon't. I'm still not over that penalty.
MuratHe didn't even touch him!
MuratAnyway. Wasn't one box scary?
The video stays on your Mac
Makas reads the video where it is. Nothing is copied or uploaded, and nothing tracks what you do.
Runs on your Mac
- Writing down every word
- Telling the speakers apart
- Finding long pauses
- Rendering the finished video
- Your projects and edits
Sent to the AI
- The transcript and the speakers' names
- The video's file name
- Your names & terms and notes