AI & ML
Image OCR (AI vision) API
GET request to /v1/image-ocr, returning json in the standard envelope.
/v1/image-ocr8 credits / call PaidAuthentication
Send your API key as a bearer token. Create one free in the dashboard.
Authorization: Bearer $TOOLSXPO_KEYParameters
Sent as query-string parameters. The values below are the examples used throughout this page.
| Name | Example value |
|---|---|
| imageUrl | https://example.com |
| prompt | Transcribe all text visible in this image, verbatim, preserving line breaks. Output only the text. |
Example request
curl "https://api.toolsxpo.com/v1/image-ocr?imageUrl=https%3A%2F%2Fexample.com&prompt=Transcribe%20all%20text%20visible%20in%20this%20image%2C%20verbatim%2C%20preserving%20line%20breaks.%20Output%20only%20the%20text." \
-H "Authorization: Bearer $TOOLSXPO_KEY"Response
Every endpoint answers with the same envelope. The result is in data (json); meta.credits reports what the call cost.
{
"ok": true,
"data": { /* json */ },
"meta": { "credits": 8 }
}Errors
Failures use the same envelope with ok: false and a machine-readable code.
| Status | Code | Meaning |
|---|---|---|
| 400 | invalid_input | A parameter failed validation. `details` lists which. |
| 401 | unauthorized | Missing or invalid API key. |
| 402 | quota_exceeded | Out of credits on this workspace. |
| 404 | not_found | No such tool. |
| 429 | rate_limited | Too many requests. Retry after `Retry-After` seconds. |
| 503 | service_unavailable | Temporarily disabled by an administrator. |
How to use the Image OCR API
Read text out of an image (imageUrl) using a Workers AI vision model. The transcription comes back in data.text.
Bashcurl "https://api.toolsxpo.com/v1/image-ocr?imageUrl=https://example.com/receipt.png" \ -H "Authorization: Bearer $TOOLSXPO_KEY"
Important: this is AI transcription, not deterministic OCR
A vision model reads the image, so it handles messy layouts and handwriting far better than classic OCR — but it can miss or alter characters. Verify anything critical (amounts, IDs, codes) before you trust it.
Options
prompt— steer the model. The default asks for a verbatim transcription; override it to, say, "Extract only the total amount" or "Return the text as JSON".
Tips & gotchas
- Use a clear, reasonably high-resolution image; tiny or blurry text hurts accuracy.
- Images are capped at ~6 MB.
- Paid tool (vision inference). Cost is per successful call.