AI background removal on Martini takes any image — generated product, character render, or uploaded photo — and outputs a clean PNG with the subject isolated on transparency. Use it to prep cutouts for compositing, ecommerce hero shots, motion graphics, or as the first step in pipelines where the subject must be re-placed against a new AI-generated environment.
Background removal segments the foreground subject — a person, product, vehicle, garment, or any well-defined object — from the rest of the image, then exports the result as a PNG with an alpha channel. Behind the scenes the model performs precise edge detection on hair, fur, fabric edges, and translucent details like glass or smoke, producing cleaner cutouts than a manual lasso or simple chroma-key would.
On the Martini canvas, background removal lives as a tool node downstream of any image source. Drop the node, wire your image in, and the cutout writes back as a transparent PNG you can use as the input to the next step. Common downstream operations include re-compositing against a new generated background, feeding the cutout into a video model that expects a clean subject, or exporting straight to your design tool.
Start with a clean source image. Strong inputs are well-lit, with the subject occupying most of the frame and visually separated from the background. Generated images from Nano Banana 2, Flux Kontext, or GPT Image 2 work well; uploaded studio photos are even stronger because lighting is controlled.
Drop a Background Removal tool node onto the canvas and connect your image source to its input. There's usually nothing else to configure — the model handles segmentation automatically. Submit the job and a transparent PNG returns to a new node within seconds.
Inspect the cutout for edge problems on hair, fur, glass, or fine fabric. If the matte is loose, regenerate the source with a slightly cleaner background separation, or run the image through a denoise/cleanup pass before retrying.
Wire the cutout into the next node in your pipeline. For ecommerce, composite against a generated background using Flux Kontext or GPT Image 2. For storyboards, drop into the export bundle. For video, feed the cutout into a motion node so the model can animate the isolated subject without competing background detail.
Generate the subject in Nano Banana 2, then strip the background for ecommerce or compositing.
View modelUse cutouts as the subject input for Flux Kontext when re-placing characters into new environments.
View modelPair with GPT Image 2 for prompt-driven background swaps once the subject is isolated.
View modelA PNG with an alpha channel — the subject sits on transparency so you can composite it against any background downstream.
Modern background removal handles hair and fur well, especially on high-resolution sources. Very fine flyaways may still be lost; if hair is critical, run an image upscale before background removal.
Most removers prefer a single dominant subject. For multi-subject scenes, crop and process each subject separately, then re-composite.
Cast shadows are usually treated as background and removed. If you need the original shadow preserved, mask it manually or generate a fresh shadow with a downstream image edit node.
Upscale before. The remover does a cleaner job when it has more pixels along the subject edges to analyse.
Chain AI Background Removal with other AI models on Martini's infinite canvas. No GPU required — start free.
Get Started Free