Automatic segmentation
Auto mode finds every object in the image (or a cropped region) with no prompt needed. Returns up to 100 segments sorted by area.
CatalogLatency ~1-3 s depending on image size and mode Async false Rate Limit Per API key Max Input 4096 x 4096 pixels (larger images are resized)
cn8/studio-segment
Send an image URL and choose a mode. In auto mode, the service detects every object in
Learn more about Image Segmentation ImagesyncInstant previewimageanalysissegmentation
Segment an image
Auto mode finds every object; hover and click the result to explore masks.
Result
Masks show up here
Every detected object gets its own colored, clickable mask.
Overview
Send an image URL and choose a mode. In auto mode, the service detects every object in the image (or in a rectangular region you specify) and returns up to 100 segments sorted by area. In prompt mode, you point at a specific object with coordinates or a bounding box and get up to 3 mask candidates ranked by confidence. Each segment comes with a mask in COCO RLE format — compact and easy to decode on any platform — plus its bounding box, area, confidence scores, and centroid. Images larger than 4096 px on a side are resized automatically. Responses are synchronous: the result comes back in the same call.
Key capabilities
Prompt-based targeting
Prompt mode targets one object using point coordinates (foreground/background labels) or a bounding box. Returns up to 3 mask candidates ranked by confidence.
COCO RLE masks
Masks come back as compact COCO Run-Length Encoding, and can be passed straight to the Inpainting service for object removal.
Region cropping
In auto mode, optionally restrict analysis to a rectangular region {x, y, width, height} of the image.
When to use it
Object removal prep
Detect objects, let users pick which to remove, then feed selected mask_rle values to the Inpainting service.
Image annotation
Auto-generate segmentation masks for labeling workflows.
Content analysis
Count objects, measure areas, and classify regions in an image.
Input & output
input
Image URL, mode (auto or prompt), optional region/points/box
JSON body
output
Array of segments with mask_rle, bbox, area, predicted_iou, stability_score, centroid
JSON
Guides & tips
Auto vs. prompt mode
- Auto mode finds every object in the image and returns up to 100 segments, sorted by area (largest first). Very small or low-confidence masks are filtered out so you get clean, usable results.
- Prompt mode targets a single object from the points or box you provide, and returns up to 3 mask candidates ranked by confidence — pick the one that best matches what you meant.
- Either way, masks come back as COCO RLE: compact to store and easy to decode on any platform.
Segment-then-inpaint workflow
- Call this service in auto mode to detect all objects.
- In your UI, let users select which objects to remove.
- Pass the selected mask_rle values to the Inpainting service (studio-inpaint).
- Result: selected objects are seamlessly removed from the image.
Region cropping
- In auto mode, pass a region {x, y, width, height} to restrict segmentation to a portion of the image. Returned bounding boxes and centroids are in full-image coordinates.
Specs
Schema
Request body
imageUrlrequiredstringPublic URL of the image to analyze
modestring · default: autoSegmentation mode: 'auto' (find all objects, default) or 'prompt' (target specific object)
enum: auto, prompt
regionobjectAuto mode only. Restrict analysis to a rectangular region: {x, y, width, height}
pointsarrayPrompt mode. Point prompts: [{x, y, label}]. label 1 = foreground, 0 = background.
boxarrayPrompt mode. Bounding box prompt: [x1, y1, x2, y2].
Response
data.image_sizeobjectwidth and height of the (possibly resized) image
data.segmentsarrayArray of segment objects
data.segments[].mask_rleobjectCOCO Run-Length Encoding of the segment mask
data.segments[].bboxarray[x, y, width, height] bounding box
data.segments[].areaintegerMask area in pixels
data.segments[].predicted_iounumberPredicted intersection-over-union (confidence)
data.segments[].stability_scorenumberMask stability score
data.segments[].centroidarray[x, y] centroid of the mask
data.segment_countintegerTotal number of segments returned
Pricing
Billed per image analyzed, regardless of segment count.
| Service | Unit | Price |
|---|---|---|
| Image Segmentation | item | 2 credits/image |
Pricing is the same for auto and prompt modes.