> ## Documentation Index
> Fetch the complete documentation index at: https://docs.origoid.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Working with images

> How to send images to OrigoID endpoints.

Several endpoints (voter ID OCR, proof of address, face match, liveness, CSF extraction with PDF) accept images. Send them as **base64-encoded strings** in the relevant field.

## Format

```json theme={null}
{
  "front": "iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAA..."
}
```

Send the raw base64 string only. Do not include the `data:image/...;base64,` prefix.

## Limits

| Constraint                    | Value                             |
| ----------------------------- | --------------------------------- |
| Maximum per image / PDF field | **12 MB** (base64-encoded length) |
| Maximum total request body    | **28 MB**                         |
| Supported formats             | PNG, JPG, PDF (where applicable)  |

Requests that exceed these limits are rejected with `INVALID_REQUEST` and `errors[].code = "PAYLOAD_TOO_LARGE"`. The `field` points either to the specific image field that exceeds 12 MB, or to `body` when the whole request exceeds 28 MB.

Multi-image endpoints (e.g. voter ID OCR with `front` + `back`, face match with `face` + `front`) share the 28 MB body cap — each image still has its own 12 MB cap.

## Recommendations for best results

These are guidelines, not hard limits. Suboptimal images are still processed; results just become less reliable.

| Recommendation                    | Why                                                   |
| --------------------------------- | ----------------------------------------------------- |
| File size under 3 MB              | Lower latency, room to spare under the 12 MB cap      |
| Resolution 1024×768 or higher     | Better OCR accuracy                                   |
| Even lighting, minimal glare      | Reduces false negatives                               |
| Full document in frame (no crops) | Required fields stay readable                         |
| JPG quality 85 or higher          | Good balance of size and detail                       |
| Compress before base64-encoding   | Base64 inflates payload by \~33% over the binary size |

If extraction fails because the image is unreadable, you receive a clear result code (`IMAGE_UNREADABLE`, `NO_FACE_DETECTED`, `DOCUMENT_NOT_IDENTIFIED`) so you can act on it.

## Privacy

Images you send are processed in memory and **never persisted**. We do not store the images or any personally identifiable information extracted from them. Only operational metadata (timestamp, latency, result code) is kept for audit purposes.

Read our full data handling policy at [origoid.com/legal](https://origoid.com/legal).