White background for a product page
The most common use: lifting an object out of its context onto white. Flux Kontext edits rather than regenerates, so the product stays exactly the same, contact shadow included.
10% off monthly plans with code NODE10, until August 31
Import a photo, describe the new setting: the AI cuts out the subject and places it on a white studio background or in a fully styled scene, without touching the product.
Real workflow, real results. Pan and zoom freely.
Use this workflowTwo instructions, two backgrounds: the same imported photo becomes a studio packshot and a marble lifestyle shot. The product does not change by a pixel.
At a glance
Clean cut-outs, believable shadows, a reinvented setting: what took an hour in Photoshop fits in one sentence here. The workflow links your imported photo (Media node) to an Image node driven by an instruction: "isolate the product on a white studio background", "place it on a marble countertop".
The model at the heart of the workflow is Flux Kontext, specialized in instruction-based editing: it changes what the instruction asks and preserves everything else, labels and reflections included. Both branches of the demo start from the same photo: the same mechanics power an e-commerce white background or a full scene.
An effective edit instruction reads like a retouch, not a new image: the guide writing a prompt that works has a section on it. And if the result should end up as a video, the Image node plugs straight into a Video node.
Import your photo
Drop your image into the Media node: product photo, portrait, item you are selling. A smartphone shot is enough, the AI does the rest.
Describe the new background
One sentence per instruction: white studio background, marble table, sunny outdoors. Add "the product stays identical" to lock the subject.
Generate and compare
Each instruction has its own Image node: generate several settings in parallel from the same photo and keep the best one.
Refine if needed
Adjust the instruction (stronger shadow, warmer light) and regenerate: only the branch concerned is re-run.
The same canvas, tuned for different needs: each variation is two or three lines of prompt away.
The most common use: lifting an object out of its context onto white. Flux Kontext edits rather than regenerates, so the product stays exactly the same, contact shadow included.
The person stays, the kitchen wall becomes a studio or a street. State the light direction of the new setting: it is the light match, not the cut-out, that makes a background believable.
Beach, mountains, a foreign city. Stay honest about the use: an invented holiday photo makes a fine campaign image and a poor family memory.
Generate the setting alone, in 16:9, in the brand's colours. It works in any video conferencing tool and comes in as many variants as you have teams.
No selection tool, no mask: the instruction describes what stays and what changes. Hair and soft edges, the nightmare of manual cut-outs, are exactly what instruction-based editing handles best.
An editing model left free will change the subject too. The instruction opens with what does not move: the subject, the pose, the clothes, the framing.
A front-lit subject dropped into a backlit setting fools nobody. Pick a background whose light resembles the original photo's.
Every pass degrades the subject a little. Go back to the original and write one complete instruction rather than stacking corrections.
Flux Kontext is trained for editing: it only touches what the instruction asks. Check small text (labels, logos), the one area worth reviewing.
JPG, PNG and WebP, imported by drag and drop or from your library. The result downloads in the node's format.
A few cents per generation with Flux Kontext, with the price shown on the button before you click.
That is the principle of instruction-based editing: Flux Kontext preserves the subject and only touches what the instruction names. Keep the original next to it on the canvas to compare.
No. Official ID photos follow strict rules and accept no generated retouching. This workflow is for commercial and personal visuals.
Each photo runs through its own node: duplicate the chain as many times as you have images, with the same instruction. The before and after stay side by side on the canvas.
Yes, as long as it is sharp. Motion blur or a badly lit subject carries into the render, editing does not create the detail that is missing.
Change a photo background with AI
Use this workflowFree account, no card required. The workflow opens pre-filled in your canvas.