v1.0 — out now for macOS 14+

Screenshot → Markdown
in three keystrokes.

SnipMD turns any screen content into editable Markdown — with built-in OCR, domain-aware AI translation, math-to-LaTeX, smart redaction, and 11 UI languages. Built for macOS, runs locally.

macOS 14+ · Apple Silicon & Intel · 3.0 MB · Developer ID signed

SnipMD Capture — Editor
// StackOverflow: hashable merge
def merge(d1, d2):
  res = d1.copy()
  for k, v in d2.items():
    if k in res and isinstance(v, dict):
      res[k] = merge(res[k], v)
    else:
      res[k] = v
  return res

Markdown · 1.2s

# hashable merge
def merge(d1, d2):
  res = d1.copy()
  for k, v in d2.items():
    if k in res and isinstance(v, dict):
      res[k] = merge(res[k], v)
    else:
      res[k] = v
  return res

See it in action

From screenshot to clean Markdown in 3 seconds

Watch a Stack Overflow answer go from Screenshot 2026-07-18.png to well-formatted Markdown with code blocks preserved — no typing, no re-formatting.

⌥⇧4  →  drag  →  editor opens, Markdown ready

Features

13 ways AI makes screenshots useful

OCR, AI tools, system-app extraction, and a 10-language code toolkit — all from one menu bar app.

Familiar Mac Shortcuts

⌥⇧3 ⌥⇧4 ⌥⇧5 ⌥⇧V ⌥⇧P — same modifier as macOS screenshot, overrides the system shortcut.

Vision-Powered OCR

25+ languages, fully offline. Apple Vision under the hood, with smart re-OCR (grayscale + 2× scaling) on low-confidence reads.

Look-Alike Fixer AI optional

Vision confuses 日/曰, 已/己, 0/O, full-width vs half-width punctuation. AI second pass fixes only the confusions — never the meaning.

Markdown That Makes Sense

A 200-line heuristic engine turns OCR output into well-formatted Markdown: headings, lists, code blocks, paragraph breaks. No more walls of plain text.

8 Providers, Domain-Aware BYO key

OpenAI, Anthropic, DeepSeek, MiniMax, Qwen, GLM, Gemini, Kimi. Auto-detects medical / legal / software / financial / academic terminology — so 线程 stays 线程, not 运程线.

Math OCR (⌥⇧P)

Snap a region with an equation. Vision-capable AI reads it and returns clean LaTeX. Paste into Markdown, Overleaf, or Notion. Solve mode also gives the result.

Ask AI About the Screenshot

Type a question — "what does this article say?", "summarize the key points" — and the AI answers from the screenshot's text. Multi-turn context, streaming.

AI-Generated Filenames

"Quarterly Revenue Dashboard 2026" instead of "Screenshot 2026-07-18 10.42". Sanitized for the filesystem. Edit the name in the save panel if you want.

Add to Reminders / Calendar / Contacts

One click extracts action items, calendar events, or business-card info. Review sheet lets you edit anything wrong, then push to Apple Reminders, Calendar, or Contacts.

Code Toolkit · 10 Languages

Explain in 1 sentence · generate runnable unit test (21+ frameworks: XCTest, JUnit 5, gtest, pytest, Jest, Vitest, xUnit, NUnit, RSpec, Kotest…) · Conventional Commits · add idiomatic doc comments.

Privacy by Default

8 built-in patterns: email, phone, credit card, IP, MAC, API key, Bearer token. "Redact & Copy" before sharing screenshots of dashboards or logs.

Mark It Up Before You Share

Arrow, rectangle, text, mosaic. Five tools in a clean top bar, plus undo. Copy the annotated PNG or the clean Markdown — your choice.

AI Summarize Built-in

Click "Summarize" and AI returns a concise summary of the OCR text, streamed. The "old reliable" feature that pairs with Q&A — summarize first, then drill in.

How it works

Three steps. Zero friction.

The shutter is silent. Markdown is ready before your finger leaves the shortcut.

01

Press a shortcut

⌥⇧3 ⌥⇧4 ⌥⇧5

Same modifier as macOS — SnipMD registers the hotkey first and overrides the system screenshot tool.

02

Drag a region

Frozen-screen overlay appears. Drag to select. Esc to cancel. The shutter is silent on region/window, with sound + flash on full screen.

03

Editor pops with Markdown

Vision OCR runs in < 1s. Optional AI correction + Markdown conversion happen in the background. Annotate, translate, redact, ask, or rename — then close.

Privacy & Trust

Your screenshots stay on your Mac.

  • OCR runs entirely on-device via Apple Vision. No bytes leave your Mac for OCR.
  • AI features only send the OCR text to the provider you configured — never the image.
  • Reminders / Calendar / Contacts writes only happen after you click "Add" in the review sheet.
  • No third-party SDKs. Source-signed with Apple Developer ID.
  • UI in 11 languages: English, 简体中文, 繁體中文, 日本語, 한국어, Français, Deutsch, Русский, العربية, Español, Português.
[Your Screen]SnipMD, locallyClipboard
                                      ↓ only if you click an AI feature
                                                      ↘
    [OpenAI / Anthropic / DeepSeek / MiniMax / Qwen / GLM / Gemini / Kimi]
                                      BYO key. Goes to the provider you chose, not us.

What's new in v1.0

Three months of features, one ship

Since v0.5 — every new AI capability, all wired up.

Math OCR (⌥⇧P) — capture a formula, get clean LaTeX back
Q&A on OCR text — ask anything, multi-turn context (last 3 turns)
AI-generated filenames — "Quarterly Revenue Dashboard 2026" not "Screenshot 2026-07-18"
AI OCR correction — fixes look-alike confusions (日/曰, 已/己, 0/O)
8 AI providers — was 2, now OpenAI + Anthropic + DeepSeek + MiniMax + Qwen + GLM + Gemini + Kimi
Domain-aware translation — auto-detects medical / legal / software / financial / academic
11 UI languages — was 5, now 11 (added fr, de, ru, ar, es, pt)
Smart re-OCR — auto retries with grayscale + 2× scaling when Vision returns low confidence
Screenshot → Reminders / Calendar / Contacts — one-click extraction with review sheet
Code toolkit (10 languages) — explain / test / commit / document across Swift, Java, C++, Python, C#, TypeScript, Go, Ruby, Rust, Kotlin
ImageNet classification — on-device auto-tagging (laptop, espresso, dashboard…)
Memory leak fixed — was 4.5 GB after a few hours, now stays under 300 MB

Pricing

Free for everyone. Pro for AI.

Cancel anytime. Pro is purely additive — every Free feature stays free forever.

Free

$0 / forever

Everything you need for screenshots on macOS.

  • Unlimited local screenshots
  • OCR + Markdown + Annotation
  • Vision-Powered re-OCR fallback
  • Smart Redaction (8 patterns)
  • ImageNet classification
  • Light / Dark / System appearance
  • Local history (last 40)
  • 11 UI languages
PRO

Pro

$4.99 / month

Or $39.99/year (save 33%). 7-day free trial. BYO AI key.

  • Everything in Free
  • AI Translation across 8 providers
  • Domain-aware terminology (medical / legal / software / financial / academic)
  • AI OCR correction (look-alike fixer)
  • Math OCR (⌥⇧P, LaTeX)
  • AI Q&A on OCR text, multi-turn
  • AI-generated filenames
  • Screenshot → Reminders / Calendar / Contacts
  • Code toolkit: 10 languages, 21+ test frameworks
  • Up to 1,000 screenshots in history
  • Custom Markdown templates
  • Priority support

FAQ

Common questions

Does SnipMD work offline?
Yes — OCR, Markdown, Annotation, Redaction, and Math OCR transcription are 100% local. AI features (translation, OCR correction, Q&A, math solve, AI naming) need internet and your own API key.
Which AI providers do you support?
8 — OpenAI, Anthropic, DeepSeek, MiniMax, Qwen, GLM, Gemini, Kimi. Bring your own key (BYOK); we never proxy through our own servers. The translator auto-detects the domain (medical, legal, software, financial, academic) and applies the right terminology.
Why ⌥⇧3/4/5/V/P instead of macOS defaults (⌘⇧3/4/5)?
macOS defaults are held at higher priority than any third-party app, so we can't override them. We picked ⌥⇧ (Option + Shift) to avoid the conflict. ⌥⇧V pastes the latest OCR result; ⌥⇧P captures a math region and returns LaTeX. Custom rebinding is on the v1.1 roadmap.
How does Math OCR work?
Press ⌥⇧P, select a region with an equation. A vision-capable AI model (MiniMax M3) reads the equation and returns clean LaTeX. Solve mode also gives the result. Paste straight into Markdown or Overleaf.
What does "AI OCR correction" do?
Apple's Vision OCR sometimes mistakes look-alike characters (日/曰, 已/己, 0/O, 1/l/I). The optional AI correction pass sends the raw OCR text back through your AI provider with a strict prompt: "fix only obvious confusions, never the meaning, preserve structure". The original text is kept if the AI makes changes that don't look right.
Does SnipMD write to my Reminders / Calendar / Contacts?
Only after you click "Add" in the review sheet — never automatically. The AI extracts a candidate list (todos, events, or contacts), you see and edit it, and only then does SnipMD ask for system permission and push to Apple Reminders / Calendar / Contacts. We never write silently.
Which languages does the Code Toolkit support?
10 source languages (Swift, Java, C++, Python, C#, TypeScript, Go, Ruby, Rust, Kotlin) across 21+ test frameworks (XCTest, Swift Testing, JUnit 5, GoogleTest, Catch2, pytest, unittest, xUnit, NUnit, MSTest, Jest, Vitest, stdlib testing, testify, Minitest, RSpec, built-in #[test], Kotest). The Add Doc Comments feature speaks all 10 source languages too.
How is this different from TextSniper?
TextSniper does OCR only. SnipMD adds AI translation (8 providers, domain-aware), redaction, annotation, code tools (10 languages), math OCR, Q&A, screenshot-to-Reminders/Calendar/Contacts, and a Markdown-first workflow.
Is there a Windows / Linux version?
Not planned. macOS-only for now — Apple Vision OCR + macOS-native shortcuts are the moat.