JevMade hello@JevMade.com
← Back to guides

JevMade field notes / Application architecture guide

Detect and skip sponsor reads with transcript and audio evidence

Sponsor Skip documents transcript-only, hybrid, and audio-only pipelines for detecting sponsor reads while deterministic code controls timestamps, jumps, confidence settings, and usage totals.

Original by trungdq88ClassificationGitHub README and Jev moduleSource reviewed

Before you dive in

What you’ll find in the original

  1. Use transcript-only mode when captions are available: Jev marks sponsor reads and code skips the bounded interval.
  2. In smart mode, use transcript boundaries first and listen only around candidate reads, with a timed transcript fallback when audio does not confirm.
  3. Keep timestamp arithmetic, jump size, thresholds, and cost accounting in code rather than asking Jev to generate playback controls.
Worth knowing

Requires a TypeSafe key; audio modes also require Deepgram and send played audio for transcription. Caption availability and transcript quality limit the default mode, while listed hourly costs are author estimates.

More guides →See the experiments →