FLUX.1 Kontext Guide: Dev, Pro, Max, and Practical Image Editing

Genoriq Team · Oct 10, 2026

You already have a photograph you like, but one detail needs changing: a jacket should be green, a room needs a different wall color, or a café sign has the wrong wording. Recreating the whole scene from a text description gives the model too many opportunities to change everything else. FLUX.1 Kontext is built around a more direct starting point: an existing image and an instruction about what to edit.

This guide explains the Dev, Pro, and Max variants available on Genoriq, then walks through three concrete editing tasks. The prompts are original examples to try with your own images, not published test results. The cover is a conceptual illustration of image editing, not a FLUX.1 Kontext benchmark sample.

1. What FLUX.1 Kontext actually does

Black Forest Labs describes Kontext as a family of models that jointly use image and text context. Its documented capabilities include local changes, scene transformations, character preservation, and editing text inside an image. “Kontext” identifies this editing family; it should not be confused with every model carrying the FLUX name. Black Forest Labs model overview.

For the user, the distinction is practical. A text-to-image prompt describes a new picture. A Kontext editing prompt points at something already visible and says how it should change. “Turn the red jacket green” has an anchor in the uploaded photograph. Describing an entirely new traveler, landscape, outfit, and camera shot can instead encourage a much broader reinterpretation.

Treat preservation as an objective to verify. A generative edit may redraw small features even when the instruction asks to keep them. For a portrait, compare the face and hands; for architecture, compare windows, railings, and perspective lines.

2. Dev, Pro, and Max: understand the names

These are different model variants, not three output sizes or a single quality slider. Their official positioning helps you decide what to investigate, but does not predict the winner for every photograph.

Dev’s architecture and editing focus are documented in the official Dev announcement. Pro and Max’s positioning comes from the model overview. These descriptions are provider claims, not a Genoriq quality ranking.

In Genoriq, all three entries currently use image-to-image mode. Start with one reference image and an edit instruction. BFL’s wider product family and other providers may expose additional modes; an option documented elsewhere is not automatically available in the Genoriq form. Check the displayed credit estimate before submitting, rather than treating a provider’s API dollar price as your Genoriq charge.

3. First edit: recolor a jacket with Dev

Choose a clear photograph of a person wearing a plain red jacket. Open FLUX.1 Kontext Dev, upload that photograph, and use this instruction:

Change only the red jacket to dark forest green. Keep the same person, facial features, hairstyle, pose, jacket cut, fabric texture, background, and lighting. Preserve the existing folds and shadows. Do not add patterns or accessories.

The word “jacket” identifies the edit target; the surrounding list names details that matter. This is more actionable than “make the outfit better.” Check the collar, cuffs, zipper, and boundary between fabric and skin. A convincing green color is not enough if the jacket’s shape has changed.

WaveSpeed’s Dev documentation lists image input and editing controls. Use the controls your Genoriq form actually presents; you do not need to reproduce every API parameter from a third-party tutorial. If the person changes noticeably, retry from the original photo with the narrower jacket instruction.

4. Scene change: use Pro to replace a setting

For a more substantial edit, try a travel portrait with one clearly visible adult subject and a simple background. Select Kontext Pro and upload the original photo:

Replace the background with a quiet pine forest on an overcast day. Preserve the person's identity, expression, hairstyle, clothing, body proportions, pose, and position in the frame. Keep the original camera distance. Adjust environmental light gently so the subject belongs in the forest. Add no other people.

This request changes the setting while explicitly anchoring the subject. The provider documents reference-based editing and successive refinements for Kontext Pro.

Review hair edges, fingers, shadows, and the transition around the shoulders. If the forest works but the face drifts, return to the original input and simplify the request. Do not repeatedly feed an increasingly inaccurate face into the next edit. Keep an approved intermediate only when it still meets your identity requirements.

5. Text edit: give Max an exact replacement

A café sign is a useful, concrete typography task. Use a photo with a short, readable sign that says “OPEN.” Select Kontext Max and request:

Replace the word "OPEN" on the café door sign with "WELCOME". Keep the lettering on the same sign, with the same perspective, letter color, and overall typographic style. Preserve the door, window reflections, frame, handle, and surrounding wall. Do not add any other words.

Kontext Max documentation emphasizes instruction following and typography. That makes lettering a relevant task to try, not a promise that every character will be correct.

Check spelling letter by letter at full size. Also inspect whether the sign’s border, glass reflections, or door geometry changed. If exact brand lettering is essential, use the edited photograph as a background and place final text with a design tool. Longer passages and tiny lettering are poor candidates for accepting a result without close inspection.

6. Choose the next step from the actual edit

Keep the original photograph beside the result. First ask whether the requested change happened. Then ask whether anything important changed unintentionally. Separate those decisions: a beautiful new forest cannot compensate for a different face.

For another attempt, change the instruction that addresses the failure: identify the target more precisely, remove competing requests, or restore an earlier input. Switching from Dev to Pro or Max is another experiment, not an automatic repair. Save the variant name, source image, prompt, and accepted output so the result remains traceable.

Start with one of these small tasks in Genoriq: a jacket recolor, a portrait background, or a short sign replacement. You will learn what a particular Kontext variant does with your image, and whether its result is ready to use.