Background Removal (Base64)#

Endpoint#

POST /v1.0/image-without-background-base64

Base URL: https://api.withoutbg.com · Spec: openapi.json

Example (cURL)#

curl -X POST \
  -H "X-API-Key: $WITHOUTBG_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"image_base64":"<raw_base64>","keep_mask_base64":null,"remove_mask_base64":null}' \
  "https://api.withoutbg.com/v1.0/image-without-background-base64"

Authentication#

Provide your API key in the request header:

X-API-Key: <your_api_key>

Request#

Content-Type: application/json

Fields#

• image_base64 (string, required) — Input image as a raw Base64 string (no data: URI prefix). Supported formats: JPEG, PNG, WebP, AVIF, HEIC, TIFF, BMP, GIF. Maximum decoded size: 20.0 MB. Invalid Base64, non-image data, unsupported formats, or oversize payloads fail with HTTP 422.
• include_image_type (boolean, optional) · default true — When true (default), include photo vs graphic probabilities in the response (image_type / X-Image-Type-*). Set false to skip.
• keep_mask_base64 (string, optional) — Optional Keep mask: marked pixels become fully opaque. One stroke on a person or object that was cut away brings back all of it. It is an image with the same width and height as the input (after EXIF orientation) or as the returned matte (processed_dimensions). White (light, opaque) pixels mark the area; black or transparent pixels mark nothing, so paint white strokes on a black or transparent canvas. The image is matted, then the strokes edit that matte: a rough stroke is enough, then edges are refined again. To fix a result further, send all strokes so far again. Where keep_mask and remove_mask overlap, Remove wins. Maximum size: 5.0 MB. Raw Base64 string (no data: URI prefix).
• remove_mask_base64 (string, optional) — Optional Remove mask: marked pixels become fully transparent. One stroke on an extra person or object removes all of it. It is an image with the same width and height as the input (after EXIF orientation) or as the returned matte (processed_dimensions). White (light, opaque) pixels mark the area; black or transparent pixels mark nothing, so paint white strokes on a black or transparent canvas. The image is matted, then the strokes edit that matte: a rough stroke is enough, then edges are refined again. To fix a result further, send all strokes so far again. Where keep_mask and remove_mask overlap, Remove wins. Maximum size: 5.0 MB. Raw Base64 string (no data: URI prefix).

Response#

Success Content-Type: application/json

Body fields#

• input_dimensions (array<integer>, required) — Decoded input size as [width, height] (pixels).
• processed_dimensions (array<integer>, required) — Size after server prep downscale (longest-side / megapixel caps) as [width, height]. Distinct from model inference resolution (~1024 long side).
• was_downscaled (boolean, required) — True when the input was downscaled during server prep.
• image_type (object, optional) — Photo vs graphic class probabilities. Present when include_image_type is true (default); omitted when false. Never returned by car endpoints.
• img_without_background_base64 (string, required) — Result image as Base64 PNG (RGBA) with transparent background. Prepend data:image/png;base64, to display in a browser.

Errors#

Related endpoints#

• Background Removal (Binary) — POST /v1.0/image-without-background