跳到正文
原文
Hacker News· cyrillsemah·· 3 小时前精选AI 评分41

Sygal:本地将 PDF、扫描件和网页转换为 Markdown

Show HN: Sygal – Convert PDFs, scans and web pages to Markdown, locally

AI 导读

Sygal 是一款本地运行的文档转 Markdown 工具,支持 PDF、Word、扫描件、网页、YouTube 链接等 29 种格式,输出 clean Markdown 与结构化 JSON。

推荐理由

本地完成 PDF 与扫描件到 Markdown 的转换,可直接接入 RAG pipeline,为不愿把敏感文件上传的处理流程提供替代方案。

正文 · 原文

Sygal

NewLaunch offer: €25 one-time, no subscription

Turn any document into clean, AI-ready Markdown,
right on your machine.

PDF, Word, Excel, images, scans, web pages, even YouTube links: Sygal converts them locally into clean Markdown and structured JSON, ready to drop into a prompt, a knowledge base or a pipeline. On scans, images and web pages in particular, that also means far fewer tokens for your AI.

14-day free trial · macOS 12+ · Windows 10+

One-time purchase · no subscription · no account

Explore the CLI

Converted in 1.7s

100% on-device

Reads everything you throw at it

The conversion

From raw to
clean Markdown.

Drop a dense PDF, a spreadsheet or a screenshot. Sygal returns clean Markdown you can read, or JSON your pipeline can parse.

Or paste a screenshot⌘V

Why the output matters

A raw file and clean Markdown aren't the same to an AI.

Sygal normalizes every document into one clean, structured output your AI can actually work with. What that changes depends on where the file comes from.

2 to 7x lighter, where it counts

Images, scans, PDFs, web & YouTube

A scan, a screenshot, a scanned PDF or a web page carries a lot your AI pays to read but never uses: pixels, layout, markup. Converted to text, the same content can run roughly 2 to 7x lighter on tokens, and the model reads what matters. YouTube links come in as clean transcripts, ready to summarize or search.

The gain is largest on visual and web content; exact figures vary with each file.

One clean, consistent output

Word, Excel, PowerPoint

Already-structured files get a different win: Sygal standardizes them into the same clean Markdown and JSON as everything else. One predictable output for your whole pipeline, ready to parse, search and reuse, with no format-by-format handling.

Whatever a document starts as, it ends the same way: clean, structured, and ready for your AI.

In practice

Drop. Convert. Review.
Export.

From a raw file to something usable, in seconds.

01

Drop

A file, a URL, or a screenshot with Cmd+V. Sygal detects the format automatically.

⌘Vto paste

02

Convert

Fully local processing. For OCR on an image or scan, the provider of your choice, with your own key.

~1.7s per file

03

Review

Native preview of the original file right next to the result. Edit the Markdown or JSON directly before exporting, nothing is locked.

before export

04

Export

Markdown, JSON, or both at once. Export right away, and the result also stays in your local library, searchable anytime.

.md/.json

What Sygal does

A full conversion engine,
not just an exporter.

Three groups of features: the essentials, what few tools offer, and what you need for daily use and volume.

The essentials

29 supported formats (PDF, Word, Excel, PowerPoint, HTML, email, EPUB...)

Clean Markdown output and structured JSON

Native preview of the original document, next to the result

Edit the result before exporting

What few tools offer

  • OCR of your choice: OpenAI, Claude, Mistral, NVIDIA, Gemini or a custom provider, with your own API key
  • Direct import from a URL
  • A persistent local library, with full-text search
  • Token count shown on every document, so you see exactly what a model reads

For daily use and volume

  • Batch conversion
  • Filters and search in the library
  • Simple retry if a file fails
  • Direct copy-paste (Cmd+V) for screenshots

Who it's for

Three ways to use Sygal.

Three profiles, one conversion engine.

Developers and AI builders

Prepare your documents for a RAG pipeline, a vector store, or an agent, Markdown and JSON ready to use directly, with no extra cleanup step.

Professions handling sensitive documents

Legal, accounting, HR, healthcare. Your contracts, payslips or case files never leave your machine, except the images you choose to send to OCR.

Everyday document use

Meeting notes, internal documentation, research. A local, searchable library of everything you've converted.

Bring your own engine

Your OCR, your rules.

Images run through the OCR provider you pick, with your own API key. Pick on cost and volume, and switch whenever you like.

OpenAIDefault

API cost

Billed per token, by image resolution.

Pick for

Everyday, mixed documents.

MistralBest value

API cost

Flat price per page, with an even cheaper async batch mode.

Pick for

Large volumes of scans.

ClaudePremium

API cost

Billed per token; the priciest at volume.

Pick for

When you specifically want Claude.

NVIDIA NIM and Gemini are also available, along with a custom provider. NVIDIA offers a free option to get started with no API budget.

OCR runs through your own API key and the engine you choose: no markup, swap anytime. Your documents are converted 100% locally and never leave your machine; only OCR on an image and importing a URL you choose reach the network.

Privacy by construction

Your documents stay on your machine.

Two columns, no ambiguity: what stays, what can leave.

Stays on your machine

  • Your files (Word, Excel, PDF, HTML, email...)
  • The text converted to Markdown and JSON
  • Your local library and its history
  • Your API keys, encrypted locally

Can go over the network, on your request

  • An image or a scan, sent to the OCR provider you choose, with your own key
  • A URL, if you import one

No account to create. No telemetry. Sygal never sees your documents.

See the full privacy policy

For your agents and pipelines

Your AI agents deserve
clean Markdown too.

The same conversion engine exists on the command line. One-command install, strict JSON output for your scripts, compatible with orchestrators like Hermes or OpenClaw.

npm install -g sygal-cli

Explore Sygal CLI

Testimonials

What the people using it have to say.

Three reviews, three different uses.

Privacy isn't negotiable in my line of work. Sygal converts my files locally, no cloud, no account. I can finally get AI-ready documents without compromising my clients' data.

LC

Laurence C.

Executive Assistant

I feed Sygal's JSON output straight into my AI pipelines. The structure is consistent every time, exactly what I needed.

TS

Thomas S.

CEO

Fast and efficient. It saves half a day of processing, for me and for the teams. Easy to use and genuinely useful.

JD

Jonathan D.

Senior Director

Pricing

One price. One purchase.
No subscription.

No usage meter, no surprise renewal.

Launch offer

€25€29

One-time purchase, launch offer

  • Licence for 2 seats, usable on the desktop app or the CLI, your choice
  • All features: conversion, OCR (BYOK), library, batch export
  • Updates included for the major version
  • 14-day free trial before purchase
  • Refund available within 14 days

Start free trial

No card required for the trial. No subscription, ever.

Already decided? Buy directly

Need several licences for your team? See the team offer

FAQ

Questions, answered.

Everything you might want to know before you try Sygal.

From the maker

“I'm an independent developer in Paris. I built Sygal for two reasons: to stop using tools that upload everything, lock you into a subscription, and still hand back messy text; and to stop burning through my AI credits. A raw document drags along a lot of noise an AI has to read without needing it. Sygal is the tool I wanted for myself: local, paid once, built to hand my AIs clean text. If it spares you the same hassle, that's the whole point.”

Sygal

Cyrill Semah

Creator of Sygal

Ready to try it

Stop copy-pasting.
Start converting.

14-day free trial. No card required. One-time purchase after that, no subscription.

macOS 12+Windows 10+No account

来源:Hacker News · sygal.app