Three steps.
Under thirty seconds.

Whether you are researching a podcast interview, following a cooking tutorial, or watching a tech review, creators mention products, books, and tools constantly. Manually tracking those mentions is exhausting. Here is exactly how Cureyt automates it — and why the architecture matters.

Step 1: Install the Chrome extension

Cureyt is a Manifest V3 Chrome extension that weighs under 200KB. It only activates on YouTube watch pages, so it never slows down browsing on other sites. Once installed, it injects a single button into the YouTube action bar — the same row that holds Like, Share, and Save. No configuration, no account, no API key.

The extension requests two permissions: access to YouTube pages (to inject the button) and activeTab (to read the transcript). It does not read your browsing history, cookies, or personal data.

Step 2: Click "Find Mentions"

When you click the button, Cureyt fetches the video transcript directly inside your browser using YouTube's public timedtext endpoints. If the auto-generated captions are unavailable, it falls back to the on-page transcript panel. This is a deliberate architectural choice: because the fetch happens from your own IP, YouTube sees a normal viewer, not a centralized scraping server.

The transcript is pre-processed before extraction: intro and outro segments are trimmed, filler words are stripped, and consecutive duplicate lines (common in auto-captions) are collapsed. This reduces noise and cuts API costs by 10–25%.

Step 3: Review categorized mentions

The cleaned transcript is sent to a tuned large language model with a single structured prompt: extract every book, tool, supplement, study, person, and product mentioned, with timestamps and confidence scores. The model runs on a 128,000-token context window, so even a four-hour podcast fits in one pass.

Results appear in a sidebar panel on the right side of the video. Each mention is color-coded by category: books in gold, tools in green, supplements in blue, studies in red, people in purple, and products in orange. Clicking a timestamp jumps the video to the exact moment the mention occurred. Clicking a Buy link opens an Amazon search with your local affiliate tag.

What types of videos work?

Cureyt was originally built for long-form podcasts, but the same extraction engine works on any YouTube video with a transcript. Here are the most common use cases:

• Podcasts and interviews — The primary use case. Extract every book recommendation, supplement stack, and tool mention from three-hour conversations with Huberman, Ferriss, or Rogan.
• Cooking and recipe videos — Pull ingredients, utensils, and appliances mentioned by chefs and home cooks. Turn a watched recipe into a shopping list.
• Beauty and makeup tutorials — Capture every product, brush, and brand recommended by creators. Build a wishlist without pausing the video forty times.
• Tech reviews and unboxings — Extract hardware, software, and accessory mentions from review channels. Compare what three reviewers recommend for the same use case.
• Fitness and workout content — Track supplements, equipment, and gear mentioned by trainers and athletes. See which protein brands and mobility tools come up most often.
• DIY and home improvement — Capture tools, materials, and supplies from project videos. Turn a weekend project plan into a complete shopping list.
• Lectures and conference talks — Extract papers, books, and references cited by academics and speakers. Build a reading list from a single ninety-minute lecture.

How the library and sync work

Every mention you save is stored in your personal library inside the extension. The library is searchable by category, keyword, and video title. You can export your saved mentions as Markdown, or sync them to a web dashboard for cross-device access.

For signed-in users, the library syncs bidirectionally between the extension and the cloud. Save a mention on your laptop, view it on your desktop, unsave it from your phone. The sync uses last-write-wins conflict resolution — simple, deterministic, and adequate for a single-user system.

Privacy and data handling

Cureyt sends only the extracted text transcript to the AI model, never audio or video. Transcripts are cached locally for seven days to avoid re-processing the same video. We do not store transcripts on our servers, sell browsing data, or use your listening habits for advertising. The extraction cost is roughly $0.017–$0.025 per video on average, which is why we can offer a free tier without venture funding.