All Capabilities
Fourteen request types and the tools around them, in four groups. Open any group for the details.
Computer Vision
View DetailsObject Detection
YOLOv4 or YOLOv4-tiny over COCO's 80 classes: a label, a confidence and a box for each object.
Face Detection
YuNet faces with five landmarks each, and an SFace embedding to tell one person from another.
Depth Estimation
A monocular depth map from Sobel gradients, rendered with the JET colormap.
Image Segmentation
K-means clustering in LAB colour space: a composite plus one mask per region.
Keypoint Detection
ORB corners or SIFT blobs, strongest first, with size, angle and confidence.
Image Processing
View DetailsResize
Fit inside a width and height, or stretch to it.
Crop
Any region, in source pixels.
Rotate
Any angle, with the canvas grown to fit and a fill colour for the corners.
Filter
Blur, sharpen, edge, emboss, enhance, equalize or normalize.
Composite
One image over another, with an offset, opacity and compose mode.
Transform
One request type that routes to resize, crop, rotate or filter.
Video & Audio
View DetailsVideo Effects
Grayscale, sepia, negate, blur, sharpen, edge, vignette or a flip, over the whole clip.
Stabilization
ffmpeg deshake, for handheld footage.
Audio Extraction
The clip's soundtrack as mp3, wav, m4a, flac or ogg.
Integration & Tools
View DetailsMCP Tools
Ten Model Context Protocol tools, so AI assistants like Claude can call the server directly.
TCP JSON Protocol
One JSON object per line over TCP: any language with a socket can be a client.
Start Processing Today
Build from source, launch the server, and start sending requests in minutes.
Getting Started Guide