trungdq88/youtube-sponsor-detection
User repository on GitHub
Detect youtube sponsor segment with live audio and transcript powered by Jev
- Forks
- 16
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Tools & CLIs
Sponsor Skip
A Chrome extension and small web app that skips YouTube sponsor reads: Jev names transcript lines or answers yes/no about what it heard, and code owns every timestamp that gets cut.
- Stars
- 114
- Language
- JavaScript
- Reading time
- 3 min
- Published by
- Community
- Author
- trungdq88
- Last verified
- Sep 2026
About this project
Skips sponsor reads on YouTube. A Chrome extension finds each read with Jev (TypeSafe's System One model) and jumps past it while you watch; a small web app does the same for a pasted link. Code owns every timestamp: Jev only ever names a transcript line or answers yes/no about what it heard. You need a TypeSafe API key. Smart and Listen modes also need a Deepgram key for speech-to-text. Clone this repo. Open chrome://extensions, turn on Developer mode, click Load unpacked and pick the extension/ folder.
Quoted from the project's README. Read the full README.
Works forJev for Document processing
Highlights
- Three modes: transcript only at under a cent per hour watched, Smart audio confirmation at about $0.05, and Listen-only streaming at about $0.46.
- Long transcripts are scanned in 80-line windows, and Jev answers over L042-style line IDs that code maps back to seconds.
- The pipeline scans windows in parallel, anchors around the winner, traces back to the first lead-in line, and repeats with the segment removed, up to six reads.
- A skip only happens above the confidence you set, 70% by default, and only at a phrase Jev is at least 80% sure is sponsor.
- Audio is captured from the video element itself, so no extra browser permission is needed and playback is untouched.
Quickstart
bash
npm install
cp .env.example .env
npm startWatch out
No license file, so reuse terms are unclear. Needs a TypeSafe API key, and Smart and Listen modes also need a Deepgram key; transcript fetching uses YouTube's private InnerTube API through youtubei.js and can break when YouTube changes things.
Questions about this entry
Short answers drawn from the linked project or page.
SponsorBlock's community labels are the ground truth for the eval, not the runtime source. This finds reads itself: Jev is asked which transcript line starts a read and which first names the sponsor, then code maps those line ids back to seconds. The eval scores the pipeline on about 20 sampled labelled videos and reports recall, precision, and start/end error.
It is built to avoid that but not perfectly. A skip only fires above the confidence you set, 70% by default, and only at a phrase Jev is at least 80% sure is sponsor. In the audio modes everything Jev hears has already played, so a jump cannot land before a read starts. The last jump in Listen mode can still overshoot by up to one step.
Transcript mode is under a cent per hour, Smart mode about $0.05, and Listen mode about $0.46. The spread is Deepgram billing per minute heard: Smart mode only listens around each detected read, while Listen mode streams the whole video. Transcript mode is also the only one that works where YouTube hands over the captions.
Transcript fetching goes through YouTube's private InnerTube API using youtubei.js, so it can break whenever YouTube changes things, and caption tracks are hidden from datacenter IPs, which means transcripts must be fetched from a home connection. The audio modes do not depend on that path, but they listen to one tab at a time and the content script only runs on youtube.com.
Run npm install, copy .env.example to .env, then npm start. A TypeSafe API key is required, and the Smart and Listen audio modes also need a Deepgram key.
They are scanned in 80-line windows, and Jev answers over L042-style line IDs that code maps back to seconds. The pipeline scans windows in parallel, anchors around the winner, traces back to the first lead-in line, and repeats up to six reads.
There is no license file, so the reuse terms are unclear. The entry flags that alongside the TypeSafe key requirement.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
X: Agentic browsing in Chrome
X: Text-to-screenshot (Malay Vasa)
X: Google's MediaPipe Decision Maker, tested in the browser
X: Playwright CLI + Jev vs Playwright MCP
More Jev resources like this
A Chrome extension that drafts reply suggestions on WhatsApp Web and Gmail using your own OpenRouter key, and quiets the YouTube feed by hiding Shorts, games and non-informational videos. It only suggests; you always press send yourself.
Jev Tools & CLIs#community#github#chrome-extension
Community
joeydash/quietlyJavaScript
joeydash/quietlyJavaScriptChrome and Firefox extension that reads a YouTube video's caption track and asks Jev for a sponsor probability per 30-second segment, painting the seek bar before the intro ends. It reports catching 77% of SponsorBlock's sponsor seconds across 23 videos at $0.0008 per video.
Jev Tools & CLIs#community#typescript#chrome-extension
Community
valentynkit/jev-skipTypeScript
valentynkit/jev-skipTypeScriptA Chrome extension that answers questions about any YouTube video by asking Jev which transcript window holds the answer: a Choice over about 30-second window ids plus a Noul for whether the video covers it at all, then it jumps to the moment.
Jev Tools & CLIs#community#github#javascript
Control a headed Chromium window by voice: one Jev request per partial transcript decides intent, target, and completeness in roughly 300 ms, while code applies thresholds and builds the action.
Jev Tools & CLIs#community#javascript#browser
A Chrome side panel that turns the current page's WebMCP tool schemas into typed Jev questions, predicting the tool call, its arguments, confidence, and latency as you type.
Jev Tools & CLIs#community#javascript#chrome-extension
Community
sdras/jev-webmcp-extensionJavaScript
sdras/jev-webmcp-extensionJavaScriptRender a live Jev meter over any video: Whisper transcribes, every sentence is scored, and ffmpeg exports a 16:9 edit with flags and a scoreboard. Presets cover debates, calls, and pitches.
Jev Tools & CLIs#community#python#video
Community
ChetasLua/jevmeterPython
ChetasLua/jevmeterPythonFrom the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.