In-page Uoink button
Button under supported video pages with helper status and fast capture.
feature inventory
Everything Uoink does.
Extraction is the on-ramp. The real product is the local corpus that grows every time you capture something worth studying.
job 1 / capture
Get content out of video and audio.
Capture from the browser, from a playlist, from a URL, or from a feed. The interface changes; the corpus shape stays stable.
Right-click capture
Context menu for links, thumbnails, and video pages.
Playlist mode
Queue up to 10 videos at a time with partial-failure tolerance.
Universal URL mode
yt-dlp backed extraction for supported web video beyond YouTube.
Podcast RSS feeds
Poll feeds and process episodes through local transcription.
Mobile queue bridge
Save to a playlist on mobile; let the desktop helper catch up later.
job 2 / organize
Turn captures into a library.
This is where Uoink stops looking like a transcript downloader. Every capture makes your local archive more useful.
SQLite FTS5 search
Search transcripts, titles, comments, metadata, and corpus text locally.
Topic folders
Route captures into folders from editable keyword rules.
Memory dashboard
Browse recent captures, job states, thumbnails, health, and actions.
Entity mentions
Find every corpus that names a person, tool, company, or topic.
Hook taxonomy
Classify openings across nine categories for creator research.
Taste calibration
Planned local memory layer that learns what you actually revisit.
job 3 / send it to AI
Use the model you already pay for.
Uoink does not bundle a model or force a chat UI. Paste anywhere, send directly, or let an agent call tools.
Paste flow.
Clipboard corpus for Claude, ChatGPT, Gemini, local models, or any editor that accepts markdown.
MCP flow.
13 local tools for capture, search, comments, hooks, health, taxonomy, and citations.
See agents ->Operator Skill.
Portable Skill instructions teach compatible agents how to use the corpus with citation discipline.
job 4 / tune the helper
Make the defaults fit your machine.
- BYO Anthropic keyEnables Hook Type, Comment Intelligence, and Entity Extraction. Stored in the OS credential vault.
- Clipboard budgetSet screenshot caps, compression, and comment limits before pasting into a model.
- Whisper model selectorChoose tiny/base/small/medium/large for podcast transcription cost and accuracy tradeoffs.
- Output folder pickerSend corpora where your work already lives, including markdown vault workflows.
why not Glasp, NoteGPT, or NotebookLM?
Transcript tools give you text. Uoink gives you a local asset.
Cloud tools can summarize. Browser extensions can grab a transcript. Generic MCP servers can fetch captions. Uoink's wedge is the combination: local corpus, screenshots, comments, hook classification, entity search, and MCP access. The archive compounds because it belongs to you.