Knockout API vs Robust Video Matting

Knockout API provides image and video processing through a browser app and API. Its image tools cover background removal and replacement, batch processing, masking, smart crop, stickers, outlines, shadows, and inpainting; the API also handles upscaling, portrait restoration, colorization, and object removal. Video background removal runs asynchronously, with job polling for progress. The browser Canvas supports layered designs with text and effects, and users can save canvases and cutouts to a free account. Developers can connect through REST, any HTTP client, or official Node, React, Python, and CLI options. The service says it manages GPU processing, endpoint updates, usage records, credentials, teams, and billing. Its privacy policy says uploaded images are processed in GPU worker memory, not retained after the response, and not used to train models. The Free plan is free, with 30 images per month and 60 requests per minute. Pay-as-you-go is $0.02 a month, with metered usage billed monthly; Plus is $10 a month. Paid tiers have no monthly cap and a 600-request-per-minute limit.

Robust Video Matting is a free project for separating people from video backgrounds. It processes frames with a recurrent neural network that carries temporal memory, and the project says it can perform matting in real time without additional inputs. A browser webcam demo runs the model live and can visualize its recurrent states. The conversion API accepts a video or image sequence and can produce a composited video or PNG sequence, optionally including alpha and foreground outputs.

Models are available for PyTorch, TorchHub, TorchScript, ONNX, TensorFlow, TensorFlow.js, and CoreML. MobileNetv3 is recommended for most uses; ResNet50 is larger and described as offering small performance improvements. The README reports GPU throughput figures, including 4K at 76 FPS and HD at 104 FPS on a GTX 1080 Ti, and 4K at 154 FPS and HD at 172 FPS on an RTX 3090 with FP16 under specified settings. The project cautions that file conversion runs much slower because its script does not use hardware video encoding or decoding or parallel tensor transfers. CoreML models require iOS 13 or later and do not support dynamic resolution. The code is GPL-3.0 licensed.