Auto-Reader · OCR API Auto-Reader · OCR API
Arabic- and Japanese-first OCR + translation API — manga-aware, and read panel-by-panel in the app — with broad Asian-script coverage and 13+ auto-detected languages · العربية واليابانية أولاً
Real API, real GPU. OCR an image, translate text, or clean up messy OCR — switch tools with one click.
Drop a PDF or images — get selectable text, Markdown, a translation, structured data (invoice, receipt, ID), or the same document back with a selectable-text layer. Runs on your free key; PDFs up to 20 pages (10 when translating).
Drop a PDF, images, or a comic archive (CBZ/CBR/CBT) here or
Need a smaller PDF, or Word / EPUB? Shrink it below, or the /v1/book API.
Every endpoint speaks JSON, returns in one round trip, and is documented in an OpenAPI spec your tools can read.
Image in, text out — boxes, confidence, reading order, and a copy-ready layout that preserves the page's line structure. lang=auto detects the script. Add format=markdown or format=chunks for RAG-ready output straight into your embedding pipeline.
A whole PDF in one call — per-page OCR and optional translation, billed by actual pages. No 50k-character minimums.
Every line translated with shared context in one round trip. format=overlay returns the page with the original text erased and the translation laid over it as selectable text.
100+ languages with business controls: formality (فصحى or casual), context, and per-key glossary — your terminology, enforced.
OCR-artifact cleanup or spelling/grammar fixes with minimal edits — line breaks preserved, nothing rephrased.
A whole book or your lecture notes in: get back a searchable PDF (text selectable, design untouched), Word, EPUB, Markdown — or a fully translated book with every page re-typeset. Handwriting in beta.
Make a scanned PDF much smaller for email while it stays visually crisp — no page limit. Adds a searchable text layer by default; send ocr=false to only compress (about 1 credit per 25 pages). A born-digital PDF keeps its selectable text either way. High-quality or Email-share presets.
Grammarly-style fixes for OCR'd notes in 100+ languages: proofread, rephrase, change tone, simplify, shorten, or summarize — minimal-surprise edits.
Summarize a scanned contract, extract fields from a receipt, classify a document — one call after OCR.
Scroll the cards with your mouse wheel →
Drop a scanned PDF (or page images). We shrink it for email while keeping the page design crisp — no page limit and no OCR by default, so it's fast and cheap (about 1 credit per 25 pages; big jobs run in the background and we show the cost first). A born-digital PDF keeps its own selectable text for free; tick Make the text searchable to OCR a scan in any language. Send a whole book if you like — if it's more than your credits cover we process as far as they go and show you the rest. Office files (PPT, Word) — coming soon.
Drop a PDF or images here or · any page count · you're only limited by your credits
Shrinking alone never runs OCR — it's fast, works on any language, and a born-digital PDF keeps its own selectable text for free. Signed in? Jobs run on your key, free daily allowance first, then credits. There's no page limit: if a file is bigger than your balance we process as many pages as it covers and tell you what's left.
Copy, paste, run. The same key works for every endpoint.
One key, every device. Fast OCR, translation, rewriting and file-shrinking wherever you work — your PC, your phone, or right inside your email flow.
Activate your free key →Export it as an env var so every call below just works.
bash: export NSK_KEY=nsk_live_xxx PowerShell: $env:NSK_KEY="nsk_live_xxx"
Any image, any language — lang=auto detects the script.
curl -s -H "Authorization: Bearer $NSK_KEY" -F "file=@scan.jpg" -F "lang=auto" https://api.auto-reader.com/v1/ocr
Per-page OCR for a multi-page document, in one call.
curl -s -H "Authorization: Bearer $NSK_KEY" -F "file=@doc.pdf" https://api.auto-reader.com/v1/document
Smaller file, still crisp. Add ocr=false to just compress (cheaper, no page limit); the default adds a searchable text layer.
curl -s -H "Authorization: Bearer $NSK_KEY" -F "file=@big.pdf" -F "quality=email" -F "ocr=false" -o small.pdf https://api.auto-reader.com/v1/reduce
100+ languages, straight JSON in and out.
curl -s -H "Authorization: Bearer $NSK_KEY" -H "Content-Type: application/json" -d '{"text":"مرحبا","target_lang":"en"}' https://api.auto-reader.com/v1/translate
Grammarly-style fixes, minimal-surprise edits.
curl -s -H "Authorization: Bearer $NSK_KEY" -H "Content-Type: application/json" -d '{"text":"...","action":"proofread"}' https://api.auto-reader.com/v1/rewrite
Share sheet a photo/PDF → Get Contents of URL (Method POST, Request Body = Form, add a file field = Shortcut Input, Header Authorization: Bearer nsk_live_…, URL https://api.auto-reader.com/v1/ocr) → Get Dictionary Value text → Copy. Android: an HTTP Shortcuts app POST does the same.
lang=ar) to skip auto-detect/v1/reduce for size (instant, no page limit — your credits are the only bound; a file past them is trimmed to what they cover with allow_partial=true) and /v1/book for big jobsIdempotency-Key header to dedupe repeats (free 24h replay)/v1/translate/batch (20% off)1 credit = 1 OCR page ≈ SAR 0.004 (~$0.001). Translation runs on our own GPU-hosted LLM and costs 7 credits a page — a fraction of DeepL, with the OCR included. Pages already in the target language bill as plain OCR.
| Per translated page (OCR included) | Free tier | Base fee | Document minimums |
|---|---|---|---|
| Auto-Reader OCR — ~$0.0075 | 500 OCR pages/day | none | billed by actual pages |
| DeepL API — ≈ $0.0375 + base fee | 500k chars/month | $5.49/mo | min 50,000 chars charged per file |
| Google Vision OCR — $1.50/1,000 (no translation) | 1,000/month | none | — |
DeepL per-page estimate assumes ~1,500 characters/page at their $25/1M characters API rate. Self-hosted, built for PDPL, and your images never train anything without your consent.
An agent can go from discovering this API to its first OCR call in one step — no email, no human in the loop.
One POST /signup — no email, no captcha — returns a working trial key. Agents onboard themselves.
A plain-text manifest tells any LLM exactly what we do, how to authenticate, and where the spec lives.
Full machine-readable spec at /openapi.json — code generators and agents build clients without reading docs.
Plug api.auto-reader.com/mcp into Claude, Cursor or any MCP client for instant OCR + translation tools — claude mcp add --transport http auto-reader-ocr https://api.auto-reader.com/mcp.
Fire a real keyed request from the browser. Get a key below, paste it here — the equivalent curl is generated for you.
Bring your own tools: OpenAPI spec · import it into Postman or any client generator · llms.txt for your agent
Email optional: with one you get 500 credits/day free, without one an instant 50/day test key. Upgrade any time — no card required.
With an email your key gets 500 credits/day free immediately — no confirmation step. Without one, an instant 50/day trial key. Images are processed transiently and not stored by default (PDPL; training storage is opt-in).
Bought credits and don’t see them? Paste the key you purchased with — the balance updates within seconds of payment.
Lost your key, or on a new device?
Prefer the app? Read books, comics and documents aloud on the go —