KontextFluxEdit an image
Focused AI image tool

Change an object’s color

Name one visible object and the new color. The prompt keeps material, highlights, reflections, and non-target colors unchanged.

One focused editThe task form is already configured for this job.
No account firstEligible visitors can submit one anonymous free edit.
Before and afterInspect the change instead of trusting a thumbnail.
Checking free AI capacity…
Private files · input ~1 hour · result ~24 hours
01 · Choose the change

Recolor object

Change one object's color without losing texture or lighting.

Recolor objectSee all edit tasks
One outputNo watermarkNo account
Review or copy the generated instructionBuilt from your fields; the server verifies it again

Change the bicycle frame to matte forest green.

02 · Source & result

KontextFlux sends only the prepared copy and edit instruction to a third-party AI provider after you submit.Privacy details

Example instructions

Describe the result in concrete visual terms.

These examples show the amount of detail the task needs. The editor adds preservation and realism instructions automatically.

Product variant

Change the bottle cap to satin black.

Vehicle

Change the car body to deep British racing green.

Interior

Change the armchair fabric to warm terracotta.

Better results

Give the model a smaller decision.

A clear source image and one unambiguous target usually provide more control than a long creative brief with several simultaneous changes.

Identify the exact object

Use a specific phrase such as the bicycle frame, the left cabinet doors, or the jacket worn by the person in front.

Describe finish as well as color

Matte, glossy, metallic, painted, dyed, or natural finishes affect highlights and reflections.

Use a hex value when consistency matters

A hex value gives the model a stronger color reference, although generated output is not a calibrated color-proofing system.

Know the limits

Review the generated details before publishing.

  • Exact brand-color reproduction cannot be guaranteed across displays and generated images.
  • Reflections and transparent materials may carry some of the original color.
  • This is a generative edit, not a selectable product-color database.
Continue the task chain

Use the next tool only when the next job is clear.

KontextFlux policy

How KontextFlux handles uploaded images, edit instructions, technical data, and short-lived results.

Updated July 30, 2026

The short version

KontextFlux is designed for one-off image edits without an account. We process the image and instruction needed to complete your request, protect the service from abuse, and delete short-lived files on a limited retention schedule.

Do not upload material that you do not have the right to process or information that you would not want sent to a third-party AI provider.

Information we process

  • The image you upload and the structured edit instruction you submit.
  • Prepared image details such as file type, dimensions, and size.
  • A random anonymous browser identifier stored in a signed, HttpOnly cookie.
  • Privacy-preserving hashes derived from network and browser characteristics for rate limiting and abuse prevention. We do not intentionally store raw IP addresses in the application database.
  • Job status, provider/model routing, error category, timing, and download events needed to operate and secure the service.
  • Cloudflare may process ordinary edge, security, and Turnstile telemetry under its own terms and policies.

Why we process it

  • To validate, queue, perform, store, and deliver the requested edit.
  • To enforce individual, shared-network, burst, and global free-capacity limits.
  • To detect unsafe or prohibited requests and investigate service failures.
  • To maintain security, reliability, and aggregate product performance without analyzing your image for advertising profiles.

AI providers

KontextFlux sends the prepared image and edit instruction to a third-party AI inference provider. The provider may route the request to an underlying model provider. Provider handling, retention, and model-improvement practices can change under their applicable agreements and policies.

KontextFlux does not claim that uploaded images are never used for provider improvement or training unless that claim is supported by the active provider agreement. Avoid uploading highly sensitive or confidential material.

Retention

  • Prepared input images are scheduled for deletion approximately one hour after upload.
  • Successful output images are scheduled for deletion approximately 24 hours after completion.
  • The normalized edit instruction is cleared from the job record after a terminal result or expiry.
  • Minimal operational records, hashes, error categories, and aggregate capacity counters may remain longer for security, abuse prevention, accounting, and reliability analysis.
  • Deletion is automated and may not be instantaneous during an outage or provider incident, but the service is not intended as permanent image storage.

Cookies and local browser data

The service uses a signed anonymous-session cookie to enforce free usage and protect private job results. The editor may also use browser memory or local storage for non-sensitive interface preferences. We do not use an account cookie or cross-site advertising cookie at launch.

Sharing

We share data with infrastructure and inference vendors only as needed to operate the service, comply with law, prevent harm, or protect legal rights. We do not sell uploaded images or edit instructions.

Your choices

  • Close or reset the editor to discard the browser's local preview.
  • Do not submit the AI job; local resize, compression, conversion, and prompt-building tools can run in your browser.
  • Contact hello@kontextflux.com about a privacy request and include the job ID if it is still visible. We may be unable to identify anonymous use after short-lived records expire.

Changes

We may update this policy when providers, features, or legal requirements change. The updated date at the top identifies the current version.