About

I listen to a bilingual language-lesson podcast where the hosts switch between French and English mid-sentence — that's the whole point, it's how the lesson works. Every generic ad-skipper I tried treated that code-switching as noise. It would either miss the ads entirely or, worse, chop into the lesson itself, and once I'd been burned by a wrongly cut sentence I stopped trusting the tool more than I trusted the ads.

That's the problem Listen Clean is built to solve. An LLM proposes where it thinks the ad breaks are, backed by the transcript it read — but a proposal is a hypothesis, not an instruction. Boundaries are snapped to real word timestamps, and a low-confidence interval is left in. A missed ad is an annoyance. A wrongly cut sentence is a broken product. I'd rather ship the first than risk the second.

The other rule I hold to: your original recording never changes. It's stored once, untouched, and every polished episode is a separate, replaceable render built from conservative cuts. If a cut turns out wrong, nothing was lost — you can always go back to the source and try again.

Listen Clean is a small, personal project, not a faceless company. If something here doesn't work the way it should, write to support@listenclean.com and I'll read it myself.

If you create or represent a podcast and want it removed from Listen Clean, email support@listenclean.com with the show name or feed URL. We’ll block its Listen Clean feeds and stop future syncing.