What is Snap-to-Explain?
Academic PDFs are full of figures that the surrounding text does not fully explain. A flowchart in a biology paper. A regression table in an economics study. A circuit diagram in an engineering textbook. Most readers skip these visuals, losing exactly the information the author considered important enough to illustrate.
Snap-to-Explain, Recitare's Snap-to-Explain, solves this by letting you draw a selection rectangle over any visual element. Recitare captures the region at 300 DPI, sends it to an AI vision model, and opens the Tutor Sidebar with a clear, structured explanation narrated aloud by Max. The whole process takes a few seconds. Your reading never stops.
It is the difference between skipping a figure and actually understanding it, without leaving the document or opening a search engine.
How does the Snap-to-Explain work?
The tool is built on a custom canvas-based selection engine (no third-party library; React 19 compatibility required writing it from scratch). Here is the exact sequence from click to explanation:
Activate the Snap-to-Explain
Click the crosshair icon in the reading toolbar. The cursor changes to indicate snip mode is active. The rest of the document dims slightly to help you focus on the target region.
Draw your selection
Click and drag to draw a rectangle over the chart, table, or figure you want explained. Corner brackets appear at the selection edges as visual guides. Release the mouse to confirm the region.
300 DPI capture + scanning animation
Recitare captures the selected region from the PDF canvas at 300 DPI using a 4.17x scale factor (300 / 72). A scanning animation plays, showing corner brackets, a sweeping scan line, and a 'Scanning at 300 DPI...' overlay, while the image is prepared and sent to the AI.
AI vision analysis
The high-resolution capture is sent to Claude Vision (Anthropic). The AI reads every label, axis, data point, and annotation in the figure and generates a structured explanation: what the visual shows, what the key takeaway is, and how it relates to the surrounding text.
Tutor Sidebar opens with narration
The Tutor Sidebar slides in from the right with your screenshot thumbnail, the structured explanation, and an animated audio waveform. Max reads the explanation aloud automatically using premium TTS so you absorb it without reading a wall of text.
Resume reading
Close the sidebar and your document TTS playback resumes exactly where it left off. The snip location is saved, and a Tutor Insight icon (amber lightbulb) marks the region for future revisits.
Why 300 DPI? Does it actually matter?
Yes, significantly. Screen resolution is 72 DPI. At that resolution, AI vision models frequently misread axis labels, confuse table columns, hallucinate values in dense figures, and fail entirely on small mathematical notation or chemical structures.
Recitare applies a 4.17x scale factor (300 / 72) when capturing the PDF canvas. This means that a figure which occupies 400 pixels on screen is captured at approximately 1668 pixels, enough for Claude Vision to read every axis tick, decimal point, and subscript accurately.
72 DPI (screen capture)
- ×Axis labels blurry or missing
- ×Small values misread
- ×Subscripts hallucinated
- ×Dense tables partially lost
300 DPI (Recitare capture)
- Full label legibility
- Decimal and subscript accuracy
- Dense tables read correctly
- Mathematical notation preserved
The higher resolution adds negligible latency (the image is still transmitted as a compressed PNG) but substantially improves explanation accuracy on the technical figures where accuracy matters most.
What is the Tutor Sidebar?
The Tutor Sidebar is a 340px slide-in panel that appears on the right side of your screen after a snip completes. It is designed to deliver the explanation without pulling you fully out of the document.
Screenshot thumbnail
Your snipped region is shown at the top of the sidebar so you always know which figure you are looking at.
Structured explanation
Claude Vision's analysis is formatted into a clear explanation: what the visual represents, what the data shows, and what the key insight is. Not a raw description, but a tutor's explanation.
Auto-narration with waveform
Max reads the explanation aloud the moment it arrives, using an american-female voice with a tutor persona. An animated waveform (32 bars, driven by requestAnimationFrame) plays while audio is active.
Add to Notes export
One click copies a rich HTML payload, screenshot image plus formatted explanation, to your clipboard. Paste directly into Notion, Goodnotes, Apple Notes, or any app that accepts rich clipboard content.
Can I revisit my snips later?
Yes. Every snip you make is persisted to your browser's local storage, keyed by document ID. Coordinates are stored as normalized 0–1 values so they remain accurate regardless of zoom level or screen size.
When you return to a document, Recitare renders amber Tutor Insight icons, lightbulb buttons, over every region you previously sniped. Click any icon to reopen the Tutor Sidebar and replay the explanation without drawing again.
Tutor Insight icons
Amber lightbulb buttons appear exactly where you previously sniped. They turn your annotated document into a persistent study resource. Every confusing figure you resolved stays resolved.
What kinds of visuals can it explain?
Claude Vision can analyze a wide range of academic figure types. The following categories work especially well at 300 DPI:
Purely decorative images (stock photos, background illustrations) produce less useful explanations. The tool is designed for information-dense academic visuals, not decorative content.
Who is Snap-to-Explain for?
STEM students
Physics derivations, circuit diagrams, reaction mechanisms: snipe the figure and Max walks you through it without breaking your reading flow.
Grad students and researchers
Results tables, methodology flowcharts, and statistical figures in journal articles get explained in plain language, with the key insight called out explicitly.
Medical and pre-med students
Anatomical diagrams, drug mechanism figures, and clinical trial forest plots are exactly the kind of dense visuals that reward a 300 DPI AI explanation.
Anyone who skips figures
If you have ever scrolled past a chart because it looked complicated, Snap-to-Explain makes it worth stopping. Two seconds to snip, ten seconds for Max to explain it.
How do you activate Snap-to-Explain?
Three steps:
- Open a PDF in Recitare (upload, browse arXiv/PubMed, or paste a URL to a PDF)
- Click the crosshair icon in the reading toolbar to activate the Snap-to-Explain
- Click and drag over the figure you want explained, and the Tutor Sidebar opens automatically with Max's explanation
No settings to configure, no prompts to write. The tool works immediately on any figure in any PDF you can open in Recitare.
Frequently asked questions
What is Snap-to-Explain?
Snap-to-Explain (also called the Snap-to-Explain) lets you draw a selection rectangle over any chart, table, diagram, or figure in your PDF. Recitare captures the region at 300 DPI, sends it to an AI vision model, and opens a Tutor Sidebar with a structured explanation, narrated aloud by Max so you never have to stop reading.
How does the Snap-to-Explain work step by step?
Activate the Snap-to-Explain from the toolbar, then click and drag to draw a rectangle over the visual you want explained. A scanning animation plays while Recitare captures the region at 300 DPI and sends it to the AI. Within seconds the Tutor Sidebar slides in with a screenshot of your snip, a structured explanation, and Max narrating it automatically. You can then resume reading or export the explanation to your notes.
Why does Recitare capture at 300 DPI instead of screen resolution?
Screen resolution (72 DPI) produces blurry, low-detail captures that cause AI vision models to hallucinate, especially on small axis labels, table values, and mathematical notation. Recitare uses a 4.17x scale factor (300 / 72) when capturing the PDF canvas, giving the AI enough pixel density to read every label accurately. The difference between 72 DPI and 300 DPI analysis quality is significant for technical figures.
What is the Tutor Sidebar?
The Tutor Sidebar, called 'The Brain' internally, is a 340px slide-in panel that opens after a snip completes. It shows your screenshot thumbnail, a structured AI explanation broken into readable sections, an animated audio waveform while Max narrates, and action buttons to resume reading or export the explanation. It closes cleanly when you are done, and TTS playback resumes automatically.
Can I revisit my previous snips?
Yes. Snap-to-Explain saves every snip location to your browser's local storage, keyed by document ID. When you return to a document, amber Tutor Insight icons (lightbulb buttons) appear over every region you previously sniped. Clicking one replays the explanation in the Tutor Sidebar without needing to draw again.
What AI model analyzes the images?
Snap-to-Explain uses Claude Vision (Anthropic) via the /api/explain-image route. Claude's vision model was chosen for its accuracy on academic figures: diagrams, flowcharts, data tables, chemical structures, and mathematical notation. The prompt is language-aware and adapts to match the document's language.
Can I export the explanation to my notes app?
Yes. The Tutor Sidebar includes an 'Add to Notes' button that copies a rich HTML payload to your clipboard. It includes the screenshot image and the formatted explanation text, ready to paste into Notion, Goodnotes, Apple Notes, or any app that accepts rich clipboard content.
Does Snap-to-Explain work with all documents?
Currently, the Snap-to-Explain works with PDFs rendered in Recitare's viewer. It captures directly from the PDF canvas at 300 DPI. DOCX, TXT, PPTX, and URL documents display as text and do not have a visual canvas to snip. For those formats, you can use Smart Cards (text selection) or Ask Max (chat) to get explanations of specific passages.