OCR Endpoint

Extract clean text from image URLs — fast, simple, and free to use.

Overview

The OCR endpoint wraps a slim Tesseract 5 CLI.
Provide public image URLs; receive an array of { url, text } objects. No keys, no persistent storage.

Key Features

JPEG & PNG up to 5 MB
Auto‑unwraps Google‑Images redirect links
Batch up to 10 images per request
note field when no text is detected

How it Works

Stream the image from its URL.
Convert to lossless PNG in memory.
Run Tesseract (page‑seg‑mode 6).
Trim whitespace and return UTF‑8 text.

Endpoint

Method	Path	Description
POST	`/api/v1/jsonsheet/ocr`	Returns plain text for every image URL

Request Payload

Content‑Type: application/json

{
  "urls": [
    "https://cdn.example.com/banner.jpg",
    "https://i.imgur.com/quote.png"
  ]
}

Rules

Max 10 URLs per call.
Accepted: .jpg, .jpeg, .png.
5 MB hard limit per image.

Typical Response

{
  "results": [
    {
      "url":  "https://cdn.example.com/banner.jpg",
      "text": "Coffee is always a good idea"
    },
    {
      "url":  "https://i.imgur.com/quote.png",
      "text": null,
      "note": "No text found. Ensure image has readable characters."
    }
  ]
}

Error Responses

Status	Message	Reason / Fix
400	Invalid body	`urls` must be a non‑empty HTTPS array
413	Image too large	Reduce file below 5 MB
429	Too Many Requests	Rate‑limit exceeded – wait ≤ 60 s
500	Tesseract error	Corrupted or unsupported image

Quick‑Start Snippets

Shell (cURL)

curl -X POST https://api.jsonsheet.com/api/v1/jsonsheet/ocr \
  -H "Content-Type: application/json" \
  -d '{"urls":["https://i.imgur.com/hello.png"]}'

JavaScript (Fetch)

await fetch("https://api.jsonsheet.com/api/v1/jsonsheet/ocr", {
  method: "POST",
  headers: {"Content-Type":"application/json"},
  body: JSON.stringify({ urls: [imgUrl] })
}).then(r => r.json()).then(console.log);

Python

import requests, json, pprint
payload = {"urls": ["https://i.imgur.com/hello.png"]}
r = requests.post(
  "https://api.jsonsheet.com/api/v1/jsonsheet/ocr",
  json=payload, timeout=30)
pprint.pprint(r.json())

C#

var body = new StringContent(
  JsonSerializer.Serialize(new { urls = new[] { imgUrl } }),
  Encoding.UTF8, "application/json");

using var hc = new HttpClient();
var resp = await hc.PostAsync(
  "https://api.jsonsheet.com/api/v1/jsonsheet/ocr", body);

Console.WriteLine(await resp.Content.ReadAsStringAsync());

Go

payload := map[string][]string{"urls": {imgURL}}
buf, _ := json.Marshal(payload)

req, _ := http.NewRequest("POST",
  "https://api.jsonsheet.com/api/v1/jsonsheet/ocr",
  bytes.NewReader(buf))
req.Header.Set("Content-Type", "application/json")

resp, _ := http.DefaultClient.Do(req)

Rate Limits

50 requests / minute
2 000 requests / day

FAQ

Do I need an API key? Nope. The OCR endpoint is completely open and stateless just POST your image URLs.
Is my picture stored? Never. Each image is streamed, processed in memory, then discarded. Nothing is written to disk except a short‑lived temp file and that is deleted before the response is sent.
What languages are recognised? Currently English. Additional .traineddata files (Spanish, French, German, Sinhala, Tamil) are on the roadmap and will roll out without breaking changes.
Hand‑written notes? Basic handwriting works sometimes but accuracy is low. For production‑grade handwriting recognition consider a specialised model such as Azure Cognitive Vision or Google Vision.
Can I OCR PDFs? Not yet. Multi‑page PDF → text is planned for Q3 2025.