Transcript evidence is the guardrail
Every boundary is anchored to the host's own words. When the evidence is weak, the interval stays in.
How it works
Listen Clean takes a podcast's RSS feed, downloads each episode, and transcribes every word with its timestamp. Proposed intervals are grounded in the host's own words, then an ad-free episode is rendered. Uncertain intervals are left in, not cut. The original recording is never overwritten. The result is a private RSS feed you add to your existing podcast app.
The pipeline, step by step
Bring a feed
Paste your podcast's RSS feed on the Add a show page. Confirm the preview and your private feed is set up immediately — automatically, not by hand.
Transcription
Each episode is downloaded and transcribed with word-level timestamps, so every proposed cut can be checked against exactly what was said.
LLM
An LLM reads the transcript and proposes ad-break intervals, quoting the transcript text that justifies each one.
The gate
Boundaries snap to the transcript's real word timestamps. A low-confidence interval is left in rather than risking a wrongly cut sentence.
Render
The conservative cut list is rendered into a new file. The original source recording is kept immutable, so a render can always be redone.
Why conservative cutting
Every boundary is anchored to the host's own words. When the evidence is weak, the interval stays in.
The downloaded source file is kept immutable. Every render is reproducible from the original plus the conservative cut list.
The output is a private RSS feed for Overcast, Apple Podcasts, or Pocket Casts — no new app to learn. Spotify doesn't support private feeds, so it's out of scope.
See it in context
Try 5 episodes free, no card required. Standard is $3.90/month when you're ready.