CleanListen: strip web-novel filler for audio

Remove vote-begging and author notes from chapters so listening isn't interrupted

Problem

Web novels imported from Qidian to WeChat Reading often start and end each chapter with long author notes: begging for monthly votes, announcing hiatus, thanking readers. Skim-reading can skip them, but audio playback cannot, so users resort to downloading pirated txt files and importing them to listen. The original poster says an agent to strip this filler would be simple and a huge UX win.

Solution

Paste or upload chapter text (or import txt/epub), and the model identifies and strips non-story paragraphs such as author notes and vote requests, outputting a clean version exportable as epub/txt for TTS apps. Every removal is reversible and highlighted to avoid cutting real plot.

Users

Heavy web-novel listeners using WeChat Reading, Ximalaya, or TTS tools; people maintaining their own audiobook libraries

MVP

Paste or upload chapters; model flags non-story paragraphs with undo; one-click clean epub/txt export; per-book rule memory; batch chapter processing

Revenue

Credits priced by processed word count; unlimited subscription; API for third-party listening tools

Why now

TTS listening is growing and platforms won't touch author content; LLMs separate story from filler accurately and cheaply, so a solo dev can ship a usable version fast