I produce a weekly interview podcast, roughly 70 minutes an episode. I've tried a few auto-clippers and the output was consistently unusable: clips that start mid-sentence, captions that mangle names, crops that cut off whoever is talking.
What's different here, if anything? Being specific because I've been burned three times.
Specific answers, since you asked specifically. Those three failure modes are exactly the ones the pipeline targets:
Clips starting mid-sentence. Cut points are chosen off the transcript at sentence boundaries, not fixed time intervals. A clip starts where a thought starts.
Mangled names and jargon. There's a glossary you can populate with recurring proper nouns, guest names, product names, technical terms. It biases transcription toward those spellings and autocorrects near-misses after the fact. For an interview show with recurring guests this matters a lot, and it's the single highest-leverage thing you can set up. Medical and technical vocabularies were the original driver for it.
Crop cutting off the speaker. Face tracking follows the active speaker through the 9:16 crop rather than center-cropping. For two-person interview framing it follows whoever's talking.
Additional things that come up for podcasts specifically: - Captions are editable *after* render, so a bad word doesn't mean re-rendering from scratch. - Custom outros can be appended automatically. - If your source has burned-in captions or a watermark from another tool, there's removal for both.
Honest caveat: it is not a human editor. On a 70-minute interview you'll get good clips and some duds, and you should skim before publishing. What it removes is the four hours of scrubbing to find the moments.
The glossary is the thing nobody else has. Every tool I've tried renders my co-host's name three different ways in one episode.
Tested on last week's episode. Got 11 clips, 7 I'd publish as-is, 2 fixable, 2 junk. That ratio is fine, it's better than my editor's first pass.
Seconding the glossary. I talk about specific supplement brands constantly and it used to be a disaster. Loaded ~40 terms in and captions went from embarrassing to fine.
Post to every connected account from one campaign, and turn long-form video into captioned vertical shorts automatically. Free plan, no card required.