Everything it does

The whole feature set,
at a glance.

A lecture-video reading workspace — every clean slide, note, highlight, screenshot, and AI answer stays pinned to the exact moment it came from. Bring your own model; it's yours to keep.

Start hereReads the board and the roomScreenshot & askThe intent routerAsk AIGenerate a visualYours to keepBring your own modelEditable promptsEverything else
The PairedY workspace: deck library, slide, notes, and the Ask-AI panel.
1
2
3
4
5
1Sidebar — your slides, decks & settings2Slide tools: select · comment · screenshot3Detailed Notes & Transcription, timestamped4Source-linked notes & AI answers5Ask about the slide — it auto-routes

Start here

Paste a link. Come back to a deck.

Give PairedY a YouTube URL (or a file you already have). It finds the slides, cleans them, reads them with a vision model, and pairs each one to what the speaker actually said — on our workers, while you keep working. No login to start.

YouTube or uploadOcclusion policyAuto captureRuns in the background
The Convert screen: paste a YouTube URL, pick an occlusion policy and capture type, and convert.

Reads the board and the room

Every slide, paired with the narration.

Per-slide OCR of text, equations, tables, and figures, fused with the timestamped transcript. Each slide carries clean Detailed Notes and the raw Transcription — with rendered math and Markdown, and every line keeps the second it came from.

Vision-model OCRMarkdown & LaTeXTimestamped transcriptEditable in place
A real slide beside its Detailed Notes and a Transcription tab, with timestamps.

Screenshot & ask

Box a figure. Ask about that.

Screenshot any region of a slide — a chart, an equation, a diagram — and it's pinned back to that exact spot with a numbered marker. Ask about it and the answer is grounded in what you boxed, not the whole slide.

Region captureNumbered pin on the slideFigure attached to the noteGrounded answer
A boxed region on the slide's chart, with a note asking what the plot shows and a grounded answer.

The intent router

One composer. It decides — and it all stacks here.

Type and press Enter: a fast, cheap model routes it to a note (no AI call), an answer, or a visual. Every note, highlight, and answer pins to the slide it came from and stacks in one panel — so a whole deck's worth of thinking lives in one place. Filter by kind or search across it all.

Enter auto-routesNote vs answer vs visualPinned to each slideStacked across the deck
The Ask-AI panel showing notes and answers from several different slides, each with its slide chip.

Ask AI

Answers pinned to the slide.

Ask about a selection or the whole slide; the answer is saved as a card linked to that passage, with rendered math & code and a provenance row showing exactly what it used — your selection, the notes, the transcript. Bring your own key; it runs from your browser.

Source-linked answersLaTeX & MarkdownProvenance chipsReply in-thread
An AI answer card showing the linked text, the question, a tag, and the answer with provenance.

Generate a visual

Turn a concept into a picture.

Ask the assistant to draw it. It generates a labeled diagram or an image from the slide's notes or a figure you captured — grounded in the lecture, saved as a card you can revisit.

Diagrams & chartsGenerated imagesFrom notes or a screenshotGrounded in the slide
An AI card that generated a labeled diagram of the next-word-prediction steps.

Yours to keep

Download it all. Or re-open a shared bundle.

Every run exports to files you own: clean slides, an editable PPTX with speaker notes, a printable handout PDF, Markdown, notes JSON, and a portable knowledge bundle. Grab the whole bundle in one click — or open a bundle someone sent you and pick up where they left off.

Download bundle (.zip)Editable PPTXPrintable handout PDFOpen a shared bundle
The Files menu listing every export — HTML, Markdown, PPTX, handout PDF, knowledge bundle — with Download and Open bundle in the sidebar.
The printable handout PDF: each slide with its notes set in the margin.

Bring your own model

Your model, your key.

Any OpenAI-compatible endpoint or OpenRouter powers the language, vision, image, and router steps — each pickable in Settings. Your key stays in your browser and is sent per request, never saved on a server. Toggle web search and visuals on or off.

OpenRouter / OpenAI-compatibleKey stays in your browserSeparate router modelTool toggles
Settings — AI & Tools: the LLM, VLM, image, router, and transcription models, key masked.

Editable prompts

Tune every prompt.

Every instruction the pipeline sends a model — slide notes, condensing the narration, concept extraction, the handout pass — is editable under Settings → Templates. Restyle the voice, then export or import your whole set as JSON. Your edits travel with the next conversion.

Every pipeline promptCollapsible & editableExport / import JSONReset to default
Settings — Templates: an editable system prompt open, with the rest collapsed.

The full list

And everything else

Grouped — every one of these ships today.

Convert

  • YouTube URL or file upload
  • Occlusion policy — inpaint or keep
  • Auto / board / slide capture
  • Choose the transcription model
  • Non-blocking — convert while you work

Clean slides

  • Boundary detection, crop & deskew
  • De-occlusion from real pixels
  • Decks and whiteboards alike
  • Duplicate slide removal
  • Illumination normalization

Notes & transcript

  • Per-slide Detailed Notes (Markdown)
  • Rendered LaTeX & code
  • Full timestamped Transcription
  • Edit notes or transcript in place
  • Jump from any line to the video

Annotating

  • Highlight text — yellow or blue
  • Screenshot a region of the slide
  • Numbered pins on the slide
  • Point comments & reactions
  • Selection popover: Highlight · Note · Ask AI

Ask AI & router

  • Intent router — note vs AI vs visual
  • Enter auto-routes; @ai forces AI
  • Ask a selection, a figure, or the slide
  • Generate diagrams, charts & images
  • Web search when you allow it

Provenance

  • Every answer shows what it used
  • Selection · notes · transcript · figure
  • Source-linked back to the slide
  • Rendered math, Markdown & code
  • Reply / follow-up in-thread

Notes panel

  • Every note pinned to its slide
  • Stacked across the whole deck
  • Filters: all · AI · notes · highlights · kept
  • Search notes, answers & tags
  • Compact ↔ expanded cards

Library

  • Every converted deck, one click away
  • Recents of past conversions
  • Delete a deck in one confirmed click
  • Open another run (.zip or .html)
  • Search slides, notes & keywords

Exports you own

  • Clean slide images
  • Editable PPTX with speaker notes
  • Printable handout PDF
  • Notes as portable JSON
  • Knowledge bundle (named after the video)

Model & prompts

  • OpenRouter or OpenAI-compatible
  • Separate LLM / VLM / image / router models
  • Bring your own key (browser-only)
  • Editable prompt templates
  • Export / import / reset prompts

Privacy & open

  • Key never saved on our servers
  • Notes stay in your browser
  • No login to start
  • MIT-licensed & self-hostable
  • Exports are yours to keep

Try it — no signup, no install.

Conversion runs on our workers — no key needed to start. Bring your own model when you want to ask questions of the result.

Open the appView on GitHub