Parsing Images

Bitparse accepts PNG and JPEG images in addition to PDFs. Each image is treated as a single page.

How It Works

1. Upload a PNG or JPEG via POST /parse.
2. The image is analyzed for text, tables, equations, figures, and charts.
3. The response contains a single entry in the pages array with page_number: 1.

File type is detected by magic bytes, not file extension. A .png file that is actually a JPEG will be processed as a JPEG.

Credit Cost

Each image costs 1 credit (one page).

Supported Formats

Other image formats (WebP, TIFF, BMP, GIF) are not supported and will return 400 Bad Request.

Examples

Parsing a PNG

curl -X POST https://api.bitparse.ai/parse \
  -H "X-API-Key: bp_YOUR_API_KEY" \
  -F "file=@whiteboard-photo.png"

import requests

resp = requests.post(
    "https://api.bitparse.ai/parse",
    headers={"X-API-Key": "bp_YOUR_API_KEY"},
    files={"file": open("whiteboard-photo.png", "rb")},
)

data = resp.json()
print(data["pages"][0]["text"])

Parsing a JPEG

curl -X POST https://api.bitparse.ai/parse \
  -H "X-API-Key: bp_YOUR_API_KEY" \
  -F "file=@scan.jpg"

import requests

resp = requests.post(
    "https://api.bitparse.ai/parse",
    headers={"X-API-Key": "bp_YOUR_API_KEY"},
    files={"file": ("scan.jpg", open("scan.jpg", "rb"), "image/jpeg")},
)

data = resp.json()
for elem in data["pages"][0]["elements"]:
    print(f"[{elem['type']}] {elem['content'][:80]}")

Example Response with Figure Extraction

When the image contains embedded figures or diagrams, Bitparse extracts them as base64-encoded PNGs:

{
  "pages": [
    {
      "page_number": 1,
      "text": "<title id=\"page1_elem0\">System Architecture</title>\n\n<text id=\"page1_elem1\">The diagram below shows the data pipeline from ingestion to output.</text>\n\n<figure id=\"page1_elem2\"/>",
      "elements": [
        {"id": "page1_elem0", "type": "title", "content": "System Architecture", "image_data": ""},
        {"id": "page1_elem1", "type": "text", "content": "The diagram below shows the data pipeline from ingestion to output.", "image_data": ""},
        {"id": "page1_elem2", "type": "figure", "content": "", "image_data": "iVBORw0KGgoAAAANSUhEUgAA..."}
      ]
    }
  ],
  "total_pages": 1,
  "processing_time_ms": 1102,
  "credits_used": 1,
  "credits_remaining": 498
}

Working with Base64 Image Data

Elements of type figure, image, or chart include base64-encoded PNG data in the image_data field. To save the extracted image to disk:

import base64

elements = data["pages"][0]["elements"]
for elem in elements:
    if elem["image_data"]:
        img_bytes = base64.b64decode(elem["image_data"])
        with open(f"{elem['id']}.png", "wb") as f:
            f.write(img_bytes)
        print(f"Saved {elem['id']}.png")

Tips

• Resolution matters: Higher-resolution images produce better extraction results, especially for tables and equations.
• Crop before uploading: If you only need a portion of the image, cropping it first reduces processing time and improves accuracy.
• One page per image: Unlike PDFs, every image is always 1 page and costs 1 credit.