Chrome ExtensionFLUX 3 Image Prompt Generator
Write prompts for FLUX 3 Image, the model Black Forest Labs released on October 1, 2026. It generates and edits on one endpoint, takes up to 10 reference images, renders up to 4k, and lets you place elements with bounding boxes. This generator follows BFL's own FLUX 3 prompting guide, and can still write FLUX.2 prompts if you pick a FLUX.2 model.
Be specific: describe who/what is in the image with details
Quick Palettes
Generated Prompt
Fill in the form and click "Generate" to create an optimized FLUX 3 Image prompt.
Tip: Be as specific as possible with your subject description. Instead of "a dog," try "a golden retriever puppy with a red bandana."
FLUX 3 Image Tips
- • FLUX 3 Image has no negative prompt, seed or guidance. Describe what should fill the space instead: "an empty promenade", not "no crowds"
- • A short prompt works because FLUX 3 expands it. Add detail only where you want control: medium first, then subject, setting, light, framing
- • Quote exact text, say where it goes and how it looks, and use \n for line breaks. Give each text block its own sentence or box
- • Tie every hex code to one named object: "the logo tile in #FF6C00", never a loose list of codes
- • Refer to uploads as "image 1", "image 2" and give each a job. Put the base image first
- • For edits: "Modify <ref_image_0> by ... Keep the rest of the image exactly unchanged, preserving ..."
- • Draft at 768sq, then re-render at 2k or 4k. Changing resolution re-generates, so the layout can shift
FLUX Image Prompt Templates
Copy-ready image briefs built around what this model actually rewards. Swap the [BRACKETED] parts for your own subject and settings, then paste straight into the tool.
Photoreal Scene
PhotographyMedium first, then subject, setting and light, which is the order BFL recommends for FLUX 3 Image.
A [MEDIUM, e.g. 35mm documentary photograph with fine natural grain] of a [SUBJECT] [DOING WHAT] in [LOCATION]. [LIGHT SOURCE, DIRECTION AND TEMPERATURE, e.g. soft window light from the left, warm on the skin, cool in the shadows]. The [MATERIAL DETAILS, e.g. brushed steel and worn oak grain] read clearly. [FRAMING, e.g. eye-level medium shot, subject on the left third], [ASPECT RATIO] frame.
Typography Poster (boxes)
TypographyFLUX 3 bounding boxes: one box per text block, exact words quoted inside desc.
A [STYLE, e.g. Swiss-style screen-printed] poster with the headline <headline_1> above a [SUBJECT] <hero_1> on a [BACKGROUND] <background_1>, and a date line <date_1> at the bottom. [{"id":"background_1","bbox":[0,0,1000,1000],"desc":"[BACKGROUND, e.g. flat cream paper in #FFFDF2 with fine grain]"},{"id":"headline_1","bbox":[60,80,260,920],"desc":"the words \"[HEADLINE]\" in heavy condensed sans-serif, all caps, [COLOUR]"},{"id":"hero_1","bbox":[300,150,820,850],"desc":"[SUBJECT AND RENDERING STYLE]"},{"id":"date_1","bbox":[860,200,940,800],"desc":"the words \"[DATE / VENUE]\" in small regular sans-serif, centred"}]Character from References
CharacterMulti-reference by position: each uploaded image gets one job.
A [MEDIUM] of the person from image 1 wearing the [GARMENT] from image 2, [POSE AND ACTION] in [SETTING]. Match the lighting and colour grade of image 3. Keep the face, hairstyle and build from image 1 exactly. [FRAMING], [ASPECT RATIO] frame.
Product on White
ProductNo negative prompt: describe what fills the frame instead of what to leave out.
A clean e-commerce studio photograph of a [PRODUCT] centred on a seamless white sweep, three-quarter view, filling roughly 70% of the frame. [MATERIAL AND FINISH DETAIL]. The [PART, e.g. cap] in exactly [HEX]. Flat even softbox lighting with a soft contact shadow directly beneath, crisp focus edge to edge, true-to-life colour.
Architectural Visualisation
ArchitectureConcrete medium and light direction instead of quality tags.
An architectural photograph shot on a tilt-shift lens of a [BUILDING TYPE] in [SETTING], [ARCHITECTURAL STYLE], built from [PRIMARY MATERIALS]. The building sits at a three-quarter angle showing two facades, with [LANDSCAPING] in front and [CONTEXT] behind under [SKY CONDITION]. [TIME OF DAY] light rakes across the facade and casts long accurate shadows. [ASPECT RATIO] frame.
Precise Edit
EditingBFL's FLUX 3 edit record: one change, then an explicit preserve list.
Modify <ref_image_0> by changing the [ELEMENT] to [NEW STATE, e.g. weathered copper with green patina]. Keep the rest of the image exactly unchanged, preserving the composition, the lighting direction, the shadows, the background and [ANY TEXT OR FACES]. The new material responds to the existing light.
FLUX Update Log
Custom output sizes
● LatestOctober 22, 2026- FLUX 3 Image now accepts exact width and height (256px per side up to about 16.8 MP, or 4 MP with reference images), overriding resolution and aspect_ratio.
- A new resolution value, match_input, keeps an edit at the size of the first reference image (up to 4 MP). Single-reference edits also follow instructions more closely, with no request changes and no price change.
FLUX 3 Image Klein
October 15, 2026- Black Forest Labs released FLUX 3 Image Klein, a 5.6B open-weight image model under the FLUX Kommunity License, on its own endpoint (/v1/flux-3-image-klein).
- It uses the same request format as FLUX 3 Image: up to 10 reference images, the same aspect ratios, resolutions up to 4k and bounding-box layout in the prompt. API prices start at about $0.008 per image at 768sq.
- Two variants: Distilled (16 steps, for fast generation) and Base (50 steps, for fine-tuning and LoRA training).
- BFL recommends running it with its prompt upsampler (flux-3-image-klein-pup). Without it, write full scene captions that cover framing, subject, background and lighting. Pick Klein in the generator and the prompt is written that way.
FLUX 3 Image
October 1, 2026- Black Forest Labs released FLUX 3 Image, the image endpoint of the FLUX 3 multimodal family, on one API endpoint (flux-3-image) that both generates and edits. The prompt decides which, and there is no mode field.
- Up to 10 reference images, output from a 768sq draft up to 4k (about 16 megapixels), and 15 fixed aspect ratios plus auto.
- New bounding-box layout control: a caption with tagged elements followed by a JSON list of boxes, for posters, packaging and precise region edits.
- No negative prompt, no prompt upsampling, no seed and no guidance. Web and image search grounding is on by default.
- BFL pricing per image: $0.041 at 768sq, $0.048 at 1k, $0.100 at 2k and $0.607 at 4k. Available in the BFL Playground and API and on fal.ai.
FLUX 3 Video 2K and 4K
September 10, 2026- The FLUX 3 Video endpoint added qhd (2560 x 1440) and uhd (3840 x 2176) output. See our FLUX 3 Video prompt generator for video prompts.
FLUX.2
Still available- Variants: klein (fastest), pro (production API), flex (precision and typography) and max (best editing consistency), plus dev as non-commercial local-only open weights.
- Up to 8 reference images on the API for pro, flex and max (4 on klein), up to 4MP output and any aspect ratio.
- Pick a FLUX.2 variant in the FLUX Model field and the generator switches to FLUX.2 rules.
How to Use the Prompt Generator
Pick the Model, Mode & Describe Your Vision
Keep FLUX 3 Image selected (or pick a FLUX.2 variant), then choose a mode: text-to-image, edit, multi-reference, or one of the bounding-box layout modes. Set the aspect ratio, resolution and colours, and describe the image starting with what matters most.
Generate Your Prompt
Click Generate. The tool leads with the medium, keeps only visible details, quotes any text with its placement, ties each hex code to one object, and rewrites exclusions as what should fill the space, because FLUX 3 has no negative prompt. In layout modes you get the caption plus the box list.
Copy & Create
Paste into the BFL Playground, the flux-3-image API or fal.ai, and set the same aspect ratio you picked here. Draft at 768sq, then render your favourite at 2k or 4k. For references, upload them in the order the prompt names them, base image first.
Frequently Asked Questions
What is FLUX 3 Image?
FLUX 3 Image is Black Forest Labs' newest image model, released on October 1, 2026. It is the image side of FLUX 3, the multimodal family that also includes FLUX 3 Video. A single endpoint handles both generation and editing, and the prompt decides which one happens. It accepts up to 10 reference images, renders from a fast 768sq draft up to 4k at roughly 16 megapixels, supports 15 aspect ratios plus auto, and adds bounding-box layout control written straight into the prompt.
Should I use FLUX 3 Image or FLUX.2?
For new work, FLUX 3 Image. It renders far larger images (about 16MP at 4k against 4MP), covers generation and editing on one endpoint, and adds bounding-box placement for layouts. FLUX.2 is still served and is cheaper, so it remains a good fit for high-volume drafts (klein), production pipelines already tuned for it (pro), typography-heavy work (flex) or identity-locked edits (max). Pick either in the FLUX Model field and the generator applies the right rules.
How does this prompt generator work?
You fill in the scene, subject, style, composition, lighting, colours, model, mode and resolution. Our AI follows Black Forest Labs' FLUX 3 prompting guide: medium first, only visible details, exclusions rewritten as what should fill the space, quoted text with placement, each hex code tied to one object, references named by position, and the official edit record. In layout modes it writes the caption plus the JSON box list for you.
How do bounding boxes work in FLUX 3 Image?
You write a normal caption that tags each element, like <headline_1> or <product_1>, then add a JSON list of rows after a space. Each row has an id, a bbox given as top, left, bottom and right on a 0 to 1000 grid from the top-left corner, and a desc. For edits, rows also say where the element comes from and whether it is kept, moved, added or removed. Boxes guide placement rather than clip it, every text block should get its own box, and very small boxes tend to fail.
How good is FLUX 3 Image at text?
Strong. Black Forest Labs shows multi-line posters and Japanese text in its launch examples, though it has not published an accuracy figure. Put the exact words in quotes with the capitalisation you want, say where they go and how they look, use \n for line breaks, and give each block of text its own sentence or box. Always proofread small print in the output.
Why no negative prompts?
FLUX 3 Image has no negative prompt field, and neither does FLUX.2. Black Forest Labs' method is to name the unwanted element, ask what should fill that space, and describe that instead. "No crowds" becomes "an empty promenade", and "not too dark" becomes "a warm, welcoming atmosphere with soft golden light". The generator rewrites your exclusion list this way automatically. For edits, a direct removal like "Remove the cat from the sofa" does work.
How do I use reference images and edits?
Refer to uploads by position, "image 1", "image 2" and so on, and give each one a job: the person from image 1, the jacket from image 2, the lighting of image 3. Put the base image first, since auto aspect ratio follows it. For edits, Black Forest Labs recommends "Modify <ref_image_0> by [your change]. Keep the rest of the image exactly unchanged, preserving [what matters]."
What resolutions and aspect ratios does FLUX 3 Image support?
Five resolution settings: 768sq for drafts, 1k as the default, 1.5k, 2k for print and hero images, and 4k at roughly 16 megapixels, which can take several minutes. Aspect ratios are 21:9, 2:1, 16:9, 3:2, 7:5, 4:3, 5:4, 1:1, 4:5, 3:4, 5:7, 2:3, 9:16, 1:2 and 9:21, plus auto, which follows image 1 and falls back to 1:1. Changing the resolution re-generates the image, so the composition can shift between a draft and the final.
How long should my prompt be?
Shorter than you might think. FLUX 3 Image expands a short prompt into a detailed description before it generates, so start with a sentence and add detail only for the things you want to control. Dense posters, layouts and multi-reference composites justify longer prompts.
Can I use HEX color codes?
Yes. A hex code sets an exact colour, but Black Forest Labs advises tying each code to one named object, for example "the logo tile in #FF6C00". The palette picker passes your codes through and the generator assigns each one to a specific surface.
What does FLUX 3 Image cost?
On the Black Forest Labs API it is priced per image by resolution: $0.041 at 768sq, $0.048 at 1k, $0.100 at 2k and $0.607 at 4k. Third-party hosts set their own prices.
Is this tool free to use?
Yes. You get 1 free prompt generation per day with no signup required. For unlimited access, sign up for a Promptslove membership which includes all AI tools and 29,000+ premium prompts.
Where can I use the generated prompts?
FLUX 3 Image prompts are ready for the BFL Playground, the flux-3-image API and fal.ai. FLUX.2 prompts also work on Replicate, ComfyUI and other hosts that serve FLUX.2. Plain-prose prompts carry over well to other image models too. The bounding-box JSON is specific to FLUX 3 Image.
Want Unlimited AI Prompt Generation?
Get unlimited access to all AI tools, 29,000+ premium prompts, courses, and resources designed to maximize your creative output.