FLUX.1 Kontext

Name: FLUX.1 Kontext
Author: Black Forest Labs

byBlack Forest Labs

Change one thing in plain language and the rest of the frame stays exactly as you shot it, with light, grain and palette intact

Edit a frame

Sculptural chrome citrus press and a blood-orange half on lavender plinths.

How FLUX.1 Kontext works

Start from a photo you already have, or from nothing. Describe the change like you would to a retoucher, attach references if you need them, and choose how much polish the render deserves.

Text prompt connected to FLUX.1 Kontext in Fuser with the generated AFTER HOURS storefront.

Write the edit as an instruction

Name the action, the target, the result and what must stay put, such as 'only change the jacket; background unchanged'.

Backpack and chevron textile references connected to FLUX.1 Kontext in Fuser beside the blended backpack result.

Attach one to four references

Drop in a source image to edit, or up to four to merge, and refer to each by description like 'the jacket from the second image'.

FLUX.1 Kontext controls in Fuser showing the Max tier, portrait aspect ratio and PNG output.

Pick Pro or Max and render

Iterate fast in Pro, then rerun the winning seed and prompt in Max for sharper lettering and tighter consistency.

What FLUX.1 Kontext is good at

Built around minimal drift: identity, pose, framing and untouched objects survive each edit, with lighting-aware blending and in-image text handled by the Max tier.

Cobalt-blue chair edited into the original apricot room while retaining the table, lamp and perspective.

Mask-free local edits

Swap a color, remove an object or relight a scene with one sentence. Shadows and reflections adjust to the change while framing and everything else survive. Pro is the fast default tier for this iteration work.

Roll-top backpack restyled with the orange, ivory and indigo chevron reference textile.

Four-reference compositing

Merge people, products and settings from up to four images into one scene. Assign each reference a role in the prompt, since conflicting references can cost likeness. Max Multi suits final deliverables once the composition is confirmed.

Illuminated AFTER HOURS sign above a dark green record-shop entrance at dusk.

Lettering that survives edits

Quote the new words and name the font and color to replace text on signs, posters and packaging. Max is the tier built for this, at about double Pro's cost (110.3 vs 55.2). Keep strings short; long paragraphs stay unreliable.

Orange messenger robot retaining its cyan eyes, turquoise satchel and burgundy scarf in a fern garden after successive edits.

Same character, next scene

Change the outfit, then the background, then the time of day, and identity holds. Feed the original image into each new edit rather than chaining many outputs, since softness builds after about five or six passes.

Made with FLUX.1 Kontext

A gallery spanning fashion, product design, music video, record sleeves and advertising, all in a clean photographic finish that takes on whatever medium the prompt names.

Short red-haired woman wearing a metallic silver coat over dark knitwear against ochre plaster.

Editorial fashion portrait with material juxtaposition

Forest-green and brushed-metal coffee grinder on a pale stone plinth.

Industrial product design hero frame

Drummer in an empty swimming pool under cobalt and magenta stage lighting.

Cinematic music-video keyframe

VELVET SIGNAL album artwork with ivory lettering and silver ribbons on oxblood.

Album artwork typography and graphic composition

Coral translucent sunglasses on pale blue plinths with a peach.

Advertising campaign still life

What people build with FLUX.1 Kontext

Used by teams who already have the picture and need one thing different: product retouchers, lookbook directors, storyboard artists, poster designers and photo editors.

E-commerce teams recoloring products

Turn one hero shot into every colorway by editing the product and leaving the table, light and angle alone. Swap, add or remove items without a reshoot or manual masking.

Fashion art directors building lookbooks

Combine a model, a garment and a backdrop from up to four references into one lookbook frame, naming the face and logo details that must be preserved.

Filmmakers keeping characters consistent

Drop the same character into a series of scenes for storyboards and keyframes, changing setting and lighting while the face, build and pose stay recognizable.

Signage and poster designers swapping text

Replace a headline, price or storefront line in an existing layout while keeping font, color and reflections. Short quoted strings work best, and Max sharpens the letters.

Photo editors retouching and restyling

Clear distractions, relight a frame or convert a finished photo to a named medium like watercolor or pencil cross-hatching, while the subject and composition stay put.

FLUX.1 Kontext specs

Plans and credits

Versions: Pro, Max
Inputs: Prompt (text), Images (up to 4 images)
Output: image
Aspect ratios: 1:1 Square, 21:9 Landscape, 16:9 Landscape, 4:3 Landscape, 3:4 Vertical, 3:2 Landscape, 2:3 Vertical
Cost: 55–110 credits per run

FLUX.1 Kontext questions

Max costs about double Pro (110.3 vs 55.2 per request) and adds stronger prompt adherence, typography and consistency. Pro is the fast everyday tier for color swaps, removals and iteration. Max suits final renders and text replacement on signs, posters and packaging, but is overkill for throwaway exploration.

Use a Multi variant, which accepts up to four reference images. Pro Multi (55.2) is the place to test a composition, and Max Multi (110.3) is for deliverables where identity and product fidelity matter. Describe each image's role in the prompt, because blends can lose likeness when references conflict.

Yes, the text-to-image variants work with no source image, in both Pro and Max. They are good for on-brand starter frames to refine with later edits, and the aspect ratio setting applies only in this mode. Output leans polished photoreal, and Max is the pick for short in-image lettering.

No. It edits from plain-language instructions with no masks, so it isn't built for pixel-exact selections. Chaining about five or six edits on successive outputs builds softness and artifacts, so feed the original image into each new edit instead. Dense paragraphs of rendered text are also unreliable.

It was created by Black Forest Labs and first launched on 29 May 2025 as the Pro and Max tiers. This node runs those Pro and Max tiers rather than the later open-weights Dev version.