Screen recorder built for course creators

The idea

Course creators are perfectionists on a deadline: one flubbed sentence in minute nine and they re-record the whole lesson. The generic screen recorder is already judged a commodity — two polished incumbents around $13/mo drew replaceable verdicts — because a pretty MP4 is where those tools stop and where a course creator’s real work begins.

The wedge is the lesson package. Record one take; the tool cuts filler and dead air via a text-based editor, then emits everything a course platform wants: chaptered video, transcript, captions, summary, and a quiz generated from what was actually said. Creators currently fake this pipeline by stacking a $13/mo recorder with a $16/mo editor and still writing quizzes by hand. One $24/mo tool that ships the whole artifact replaces the stack and the busywork.

Who pays

The buyer is the independent course creator selling on Teachable, Kajabi, Podia, or Skool — someone whose course revenue makes $24/mo invisible next to hours saved per lesson. Corporate L&D teams are a later up-market tier. Distribute through creator communities on Skool and Circle, YouTube tutorials on course production, and affiliate deals with course-platform ecosystems.

MVP scope

• In-browser screen plus camera recording with upload pipeline
• Word-level transcript with suggested cuts for filler, silence, and false starts
• Text-based editing where deleting words cuts the video, rendered by ffmpeg
• Auto-chapters, SRT captions, lesson summary, and generated quiz
• MP4 export plus artifact zip and unlisted YouTube upload
• Skip for v1: multi-track editing, team workspaces, AI voice correction, direct Teachable API publishing

Build LessonCut, a web-based recorder that turns one imperfect take into a finished course lesson. Stack: Next.js (App Router) with in-browser capture via getDisplayMedia and getUserMedia, ffmpeg (server-side workers on Railway or Fly) for cutting and rendering, Whisper API for word-level transcription, Claude API for chapters and quiz generation, S3-compatible storage.

Core flow: creator records screen plus camera bubble in the browser. On stop, the recording uploads and processes: Whisper produces a word-timestamped transcript; the app detects filler words (um, uh, like, you know), silences over 1.5s, and false starts, and marks them as suggested cuts. The editor is text-based — the creator reads the transcript, unchecks any cut they want kept, deletes whole sentences to remove them from the video, and ffmpeg renders the clean cut with crossfades at cut points so audio does not pop.

Then the package generates: auto-chapters with titles from topic shifts, a cleaned transcript, an SRT file, a lesson summary, and a 5-question multiple-choice quiz with answers drawn from what was actually said. Export as MP4 plus a zip of artifacts, with direct upload to YouTube (unlisted) for Teachable/Skool embedding.

Non-obvious details: render cuts with 3-frame audio crossfades to avoid clicks; cap suggested cuts at natural clause boundaries using transcript punctuation so edits never clip mid-word.

Pricing: free with watermark and 10-minute cap, $24/mo unlimited. Stripe billing.