How to Use Flux Kontext Max: Instruction Editing at 10 Credits an Image

Flux Kontext Max is the top tier of Black Forest Labs’ Kontext line on VdoBloom, it costs a flat 10 credits per image in both text-to-image and editing mode, and its defining trick is instruction editing: it changes what you ask for and leaves everything else alone. It outputs in six aspect ratios from 21:9 down to 9:16, accepts prompts up to 3,000 characters, and in editing mode takes exactly one reference image. It rates 4.5 with VdoBloom users — the higher of the two Kontext tiers — and carries the MODERATE content label.

What Flux Kontext Max is best at

The Kontext design point is preservation. Give it an image and one natural-language instruction — change the jacket to olive, swap the headline text, move the light source to the left — and it rewrites that and only that. Composition, face, product geometry and background survive. That is a genuinely different behaviour from generators that re-roll the whole frame every time you touch the prompt, and it is why Kontext is the model you reach for when ninety percent of an image is already right.

Max sits above Kontext Pro on output quality, with the lab publicly emphasising typography and prompt adherence at this tier. In practice that shows up in three places: rendered text inside the image holds together at close inspection, fine material detail survives the edit, and long multi-clause instructions get followed more often than paraphrased. It is also the tier VdoBloom’s Studio agent picks automatically for artistic and stylised looks and for premium single-image edits, and it is offered in the Story Generator as the best-quality frame model at 10 credits per frame.

Real specs and credit cost

Credits convert cleanly: Lite is $15 for 300 credits, so one credit is about $0.05. On annual Pro — $49/month for 2,000 credits — the same credit is about $0.0245. Flux Kontext Max charges the same 10 credits whether you generate from scratch or edit an upload, which makes it one of the easier models to budget.

Spec	Flux Kontext Max	Flux Kontext Pro
Credits per image	10	7
Cost at $0.05/credit	$0.50	$0.35
Cost at annual Pro rate	about $0.245	about $0.17
Modes	Text-to-image and image editing	Text-to-image and image editing
Aspect ratios	21:9, 16:9, 4:3, 1:1, 3:4, 9:16	21:9, 16:9, 4:3, 1:1, 3:4, 9:16
Reference images in edit mode	Exactly 1	Exactly 1
Prompt limit	3,000 characters	3,000 characters
Output formats	JPEG or PNG	JPEG or PNG
VdoBloom user rating	4.5	4.3
Content label	MODERATE	MODERATE

Two operational details catch people out. First, the one-image rule in edit mode is enforced before anything is sent to the provider — hand it two uploads and the request is rejected with a clear error rather than silently ignoring the second file, so no credits are wasted. Second, the safety filter runs tighter on editing than on generation; the same prompt that passes as a fresh generation can be refused when applied to an uploaded photo. That is by design, not a bug.

Non-English prompts are auto-translated before generation, so you can write your instruction in your own language without hand-translating it first.

How to prompt it well

Kontext responds to instructions, not to descriptions. The difference matters more than any other prompting tip for this model.

Name the change, not the scene. “Replace the blue backdrop with warm grey seamless paper” outperforms a fresh 200-word description of the whole photo. If you re-describe everything, you invite the model to reconsider everything.

One instruction per run. Chaining “make it night, change her coat to red, and add falling snow” into one prompt usually lands two of three. Three sequential 10-credit runs cost 30 credits and land all three, with an inspectable result at each step.

Quote text exactly. For typography work, put the literal string in the prompt — change the headline to read SPRING DROP 2026 — and state the placement. This is the tier built for that job, so give it exact copy rather than a description of the copy.

State what must not change. Adding a short clause such as “keep the pose, lighting and background unchanged” measurably tightens preservation on aggressive edits.

If your source photo shows a real, identifiable person, you must have that person’s consent before uploading and editing their image.

When to pick a different model

Flux Kontext Max is not the answer to everything, and at 10 credits it is a mid-priced model rather than a cheap one.

Pick Flux Kontext Pro at 7 credits when you are iterating. Identical capability and identical aspect ratios, thirty percent cheaper per attempt; find the right instruction on Pro, then run the final on Max.

Pick a multi-image editor when the job needs more than one reference. The hard one-image limit is the real constraint here — compositing a product into a scene, or blending a style reference with a subject, needs something like Seedream 4.5 Edit at 7 credits or Nano Banana Pro at 18 credits for 1K and 2K, 24 for 4K.

Pick a removal tool when you are subtracting. Deleting an object cleanly is a mask job, not an instruction job — FLUX Erase at 10 credits paints out the region and rebuilds it from context instead of re-interpreting your sentence.

Pick a cheaper generator for volume text-to-image. If you need forty concept thumbnails, 10 credits each adds up fast; the lower-cost generators on the platform exist precisely for that pass.

A worked example

A packaging comp. Start on the image generation page, select Flux Kontext Max, set 4:3, and prompt for the base scene: a matte black coffee pouch standing on pale concrete, soft window light from the left, shallow depth of field. Ten credits.

The shape is right but the label is placeholder. Move to the edit page, upload that single render, select Flux Kontext Max, and instruct: change the front label to read MIDNIGHT ROAST, single line, centred, in a clean sans-serif; keep the pouch shape, lighting and background unchanged. Ten credits — this is the tier’s strongest use case.

One more pass to warm the light by a stop, another ten credits, and the comp is done at 30 credits total: about $1.50 on Lite, or roughly $0.74 at the annual Pro rate. Full tiers are on the pricing page; new accounts start with 10 free credits and no card, which is exactly one Kontext Max image to try the model before committing.

Frequently asked questions

How much does one Flux Kontext Max image cost?

Ten credits, the same for text-to-image and for editing. That is $0.50 at the Lite rate of about $0.05 per credit, or roughly $0.245 per image on the annual Pro tier at 2,000 credits.

What is the difference between Kontext Max and Kontext Pro?

Same capabilities, same six aspect ratios, same 3,000-character prompt limit and the same one-image edit rule. Max is the premium tier, positioned for higher output quality and stronger typography, rated 4.5 against Pro’s 4.3, and costs 10 credits against Pro’s 7.

How many images can I upload when editing?

Exactly one. The limit is enforced up front, so a second upload returns an error rather than being ignored. Multi-reference work needs a different model.

Is Flux Kontext Max good at text inside images?

Typography is this tier’s headline strength, which is why it suits posters, packaging and ad comps. Put the exact string in the prompt in capitals or quotes and state where it goes.

What resolutions and aspect ratios does it support?

Six aspect ratios — 21:9, 16:9, 4:3, 1:1, 3:4 and 9:16 — with JPEG or PNG output. There is no separate resolution selector on this model; the ratio choice sets the frame.

What content will it refuse?

It carries the MODERATE label: standard filtering rather than a strict lockdown, and the filter is tighter on edits of uploaded photos than on fresh generations. Mainstream creative work passes; explicit and illegal content is blocked everywhere on VdoBloom. If a defensible prompt keeps getting refused, the more flexible model tiers listed on the model page are the place to look next.