This is the one-page reference for prompting FLUX.2 by Black Forest Labs. Unlike Midjourney, Flux takes no --parameters in the prompt: it reads full descriptive sentences and you tune the rest through API fields — guidance, steps, width, and height. Below is the Subject + Action + Style + Context formula and copy-paste tables for every tier, setting, and modifier that reliably moves the output.
New to Flux? Start with the 40 best Flux prompts roundup, then use this sheet while you build your own. For a deeper walk-through, see how to prompt Flux for photorealism, and grab fill-in-the-blank Flux templates when you want a running start.
The prompt formula
A strong FLUX.2 prompt is one descriptive paragraph built from four factors in this order: Subject + Action + Style + Context. Word order matters — put the most important element first, because the model weights earlier words more heavily. Lead with the main subject, then the key action, then the critical style, then the essential context and any secondary details.
- Subject — who or what, described concretely (age, clothing, material, expression).
- Action — what the subject is doing, even if subtle ("mid-stride", "pouring coffee").
- Style — photoreal, cinematic, 3D render, flat vector, and so on — one clear direction.
- Context — setting, lighting, lens, mood, and any brand colors or text.
Length: short 10–30 words for quick concepts; medium 30–80 words is the sweet spot; long 80+ words only for complex scenes. No negative prompts — FLUX.2 does not support them, so describe what you WANT, not what you don't (write "clean seamless white background", not "no clutter").
Skeleton to copy:
[Main subject, described concretely] [action], in the style of
[style/medium], [setting and context], lit with [lighting setup],
shot on [camera and lens], [color palette or hex codes], [any "quoted text"].The same skeleton, filled in:
A weathered fisherman in his sixties in a knitted grey sweater, looking just
off-camera on a misty harbour wall at dawn, photorealistic editorial style
with natural skin texture, soft golden-hour backlighting and gentle rim light,
shot on Sony A7R V, 85mm f/1.4 with shallow depth of field.Why it works: the subject leads, the action and style come next, and the context — setting, lighting, and real camera language — fills out a single sentence a photographer could shoot, which is exactly what FLUX.2 was trained to reward. It lands inside the 30–80 word sweet spot with no negatives.
Model tiers
FLUX.2 ships in five tiers. Only [flex] exposes the Steps and Guidance sliders for manual control; [pro] and [max] fix those to optimal internal values and lead on text; [klein] is sub-second with fixed params; [dev] is the open-weights build for local use.
| Tier | Best for | Notes |
|---|---|---|
| FLUX.2 [pro] | Top overall quality and best text | Fixed optimal steps/guidance; accepts JSON-structured text prompts |
| FLUX.2 [max] | Highest fidelity and typography | Fixed params; reach for it on the most demanding detail and lettering |
| FLUX.2 [flex] | Manual control | The only tier that exposes Steps and Guidance sliders to tune by hand |
| FLUX.2 [klein] | Speed and drafts | Distilled, sub-second, fixed params — passing steps/guidance has no effect |
| FLUX.2 [dev] | Local / offline | Open weights, runs locally in ComfyUI and similar |
Where to run them: the bfl.ai playground, the FLUX API, and partner platforms (fal, Replicate, Freepik, LTX, Scenario). The [dev] weights run locally in ComfyUI.
Parameters & settings
Flux has no prompt flags — you set these as API fields (on [flex] for guidance and steps). The defaults are tuned well; change them only with a reason.
| Setting | Default / range | What it does |
|---|---|---|
| Guidance scale | 2.5 default; 5–10 literal | Higher = stricter, more literal adherence (product/technical); lower = more creative freedom. Adjustable on [flex] only. |
| Steps | 30 default (15–20 drafts, up to 100) | More steps = more detail and time. Adjustable on [flex]; ignored on [klein]. |
| prompt_upsampling | on / off | Auto-enhances a short prompt with detail while keeping your intent. |
| Multi-reference | up to 10 images | Combine product + face + style refs into one output for consistent characters and brand. |
| JSON text prompts | [pro] | Specify exact text positions, fonts, and styling for precise design work. |
| Width & height | multiples of 16 | Sets both resolution and aspect ratio — there is no --ar flag. |
Aspect ratios & resolution
Aspect ratio is set by the width and height you request — there are no flags. FLUX.2 outputs up to 4MP (e.g. 2048×2048); dimensions must be multiples of 16, and 2MP (~1440–1536px on the long edge) is recommended for most work. Common ratios and starting sizes:
| Aspect ratio | Dimensions (px) | Typical use |
|---|---|---|
| 1:1 | 1440×1440 | Profile pics, album art, Instagram grid, product tiles |
| 3:2 | 1536×1024 | Classic photography, editorial, prints |
| 2:3 | 1024×1536 | Book covers, posters, Pinterest pins |
| 4:3 | 1440×1088 | Presentation slides, classic monitors, blog headers |
| 3:4 | 1088×1440 | Portraits, mobile-first cards, catalogue shots |
| 16:9 | 1536×864 | YouTube thumbnails, hero banners, desktop wallpaper |
| 9:16 | 864×1536 | Stories, Reels, TikTok, phone wallpapers |
Resolution note: push toward 4MP (2048×2048) for anything you will print or crop into; stay near 2MP for the web — it is faster and cheaper. Always keep both dimensions divisible by 16.
Lighting modifiers
Lighting sets the mood faster than any other factor. Name a real setup and FLUX.2 renders it convincingly. Drop one phrase into the context part of the formula.
| Modifier | Effect |
|---|---|
| Three-point softbox setup | Clean, even, flattering studio light — the safe default for portraits and products |
| Chiaroscuro, harsh high contrast | Dramatic light-and-shadow, deep blacks, painterly and moody |
| Golden-hour backlighting | Warm glow and long shadows, halo around edges, dreamy and cinematic |
| Soft overcast diffusion | Flat, shadowless, gentle light — natural skin, honest product colour |
| Rim light | Bright edge that separates the subject from a dark background |
| Rembrandt lighting | Small triangle of light on the shadowed cheek — classic portrait look |
| High-key lighting | Bright, airy, low-shadow — beauty, e-commerce, clean and optimistic |
| Low-key lighting | Mostly dark with selective highlights — noir, luxury, tension |
| Neon / practical lights | Coloured urban glow (magenta, cyan) reflecting on wet surfaces |
| Hard direct sunlight | Crisp, high-contrast shadows — bold fashion and street work |
| Candlelight / firelight | Warm, flickering, low and intimate — cosy and romantic scenes |
Camera & lens modifiers
Real camera and lens language tells FLUX.2 how the scene should compress, blur, and frame. Name real bodies, lenses, and film stock — specific beats generic ("professional photo" is weak).
| Modifier | Effect |
|---|---|
| Shot on Sony A7R V, 85mm f/1.4 | Flattering portrait compression with creamy background blur |
| 35mm | Natural, documentary, street perspective close to how the eye sees |
| 24mm wide-angle | Expansive scenes, interiors, landscapes — some edge stretch |
| Macro 100mm | Extreme close-up detail — jewellery, food, textures, insects |
| Kodak Portra 400 | Warm, soft film-stock colour with gentle grain — editorial and portrait |
| Shallow depth of field | Sharp subject, soft background — isolates and draws the eye |
| Deep depth of field | Everything sharp front to back — landscapes, architecture |
| Bokeh | Soft round out-of-focus highlights behind the subject |
| Drone / aerial shot | Top-down or high overhead view for scale and pattern |
| Low angle / worm's-eye | Looking up — makes the subject tower and feel heroic |
| Tilt-shift | Selective focus band that makes real scenes look miniature |
Style & medium modifiers
The style factor decides whether you get a photo, a render, or an illustration. Name one clearly and put it early; mixing three fights the model.
| Modifier | Effect |
|---|---|
| Photorealistic | Looks like a real photograph — natural texture, real lighting physics |
| Cinematic | Film colour grade, wide-frame drama, moody contrast and atmosphere |
| 3D render (Octane / Blender) | Clean CGI with glossy materials and studio reflections |
| Watercolour | Soft washes, bleeding edges, visible paper texture |
| Isometric | Angled 3D-tile look for scenes, rooms, and game-style diagrams |
| Flat vector illustration | Bold shapes, limited palette, no gradients — icons and web art |
| Film grain, 35mm analog | Grainy, slightly faded, nostalgic film-camera aesthetic |
| Product studio | Seamless backdrop, crisp reflections, e-commerce-ready lighting |
| Anime / manga | Cel-shaded characters with expressive lines and flat colour |
| Line art / sketch | Pen or pencil linework, minimal or no colour |
| Oil painting | Visible brushstrokes, rich impasto texture, classical feel |
Text & typography rules
FLUX.2 has best-in-class in-image typography, but only if you follow a few rules. For posters and precise design, use [pro] or [max], and on [pro] you can pass JSON-structured prompts that pin exact positions and fonts. Grab ready-made poster skeletons from the Flux prompt templates.
| Rule | Why |
|---|---|
| Wrap the exact words in "quotes" | Tells the model precisely which characters to render — no paraphrasing |
| Name the font and weight | "bold condensed sans-serif" or "elegant serif" guides the letterforms |
| Give placement | "centered near the top" or "lower third" lets the model plan layout early |
| Use hex color codes | Put e.g. #FF6B35 in the prompt and FLUX.2 reproduces that exact brand colour |
| Keep strings short | 1–4 words render cleanly; long paragraphs are more likely to garble |
| Use [pro] / [max] for heavy text | Best-in-class typography; [pro] accepts JSON for exact text positioning |
Text example prompt:
A minimalist gym poster with the bold headline "NO EXCUSES" in a heavy
condensed sans-serif, centered near the top, in energetic editorial poster
style, a lone runner in silhouette on an empty track at dawn backlit by
golden-hour sun with long shadows, high-contrast, deep charcoal background
with #F5C518 accent type. Rendered on FLUX.2 [pro].Why it works: the exact words are quoted and kept to two words, the font and placement are named, and a hex accent locks the brand colour — run it on [pro] or [max] for the crispest lettering.
Copy-paste example prompts
Five complete prompts that assemble the modifiers above. Paste any into the Flux playground or API and tweak the details. For more, browse the best Flux prompts.
1. Product hero shot
A matte-black ceramic coffee mug with steam rising, centered three-quarter
view on a wet slate surface, photorealistic product studio style with crisp
reflections and a seamless charcoal backdrop, dim minimalist studio lit with
a three-point softbox setup and subtle rim light, shot on 100mm macro with
shallow depth of field. Guidance 6, 1:1 at 1440×1440.Best for: e-commerce listings and ads — higher guidance and a macro lens make the material read as premium.
2. Cinematic portrait
A confident woman in her thirties with short curls and a tailored linen
blazer, looking directly at camera, photorealistic cinematic portrait with
natural skin texture, in a sunlit loft with a soft blurred window behind her,
lit with Rembrandt lighting and soft overcast fill, shot on Sony A7R V, 85mm
f/1.4, shallow depth of field, Kodak Portra 400. 3:4 at 1088×1440.Why it works: subject and framing lead, then style and real lens language — see the photorealism guide for skin and lighting deep-dives.
3. Brand poster with quoted text
A retro travel poster with the headline "VISIT MARS" in a bold geometric
sans-serif across the top, mid-century illustration style with a warm limited
palette, a stylised red desert landscape and a distant domed colony under a
pink sky, headline in #FF6B35 on a cream banner. Rendered on FLUX.2 [pro],
2:3 at 1024×1536.Best for: posters and ads that need exact words and an exact brand colour — quoted text plus a hex code on [pro].
4. Flat vector illustration
A friendly flat vector illustration of a person watering a large houseplant
in a cozy apartment, centered composition, modern flat illustration style with
bold clean shapes and no gradients, warm limited palette of terracotta, sage,
and cream, soft even lighting, for a web hero. 16:9 at 1536×864.Best for: landing-page graphics and blog headers where a clean, on-brand illustration beats a photo.
5. Hex-color brand scene
A single running shoe floating above a soft shadow, dynamic hero angle,
photorealistic product studio style, on a seamless background in exact brand
colour #0EA5E9, accents in #FFFFFF, lit with high-key studio lighting and a
gentle rim light, shot on 85mm with shallow depth of field. Guidance 7,
3:2 at 1536×1024, rendered on FLUX.2 [max].Best for: brand campaigns that must hit an exact palette — the hex codes lock the colours and [max] holds the detail.
Frequently Asked Questions
Does Flux take --parameters like Midjourney?
No. FLUX.2 has no --ar, --style, or --v flags. It reads full descriptive sentences, so you write requirements in plain English and set the rest through API fields — guidance, steps, width, and height. State the aspect ratio in words if you like, but there is no flag syntax; tag-soup and Midjourney-style flags hurt results.
What guidance scale should I use in Flux?
The default guidance scale is 2.5, which balances prompt adherence with natural, creative output. Raise it to 5–10 when you need strict, literal adherence for product, technical, or text-heavy work; lower it for more creative freedom. Guidance is only adjustable on FLUX.2 [flex] — on [pro] and [max] it is fixed to optimal internal values, and on [klein] passing it has no effect.
What is the maximum resolution FLUX.2 can output?
FLUX.2 outputs up to 4MP, for example 2048×2048. Output dimensions must be multiples of 16, and aspect ratio is set by the width and height you request rather than a flag. For most work, 2MP (around 1440–1536px on the long edge) is the recommended sweet spot — it is faster and cheaper while staying sharp.
Which Flux tier is best for text and typography?
Use FLUX.2 [pro] or [max] for anything text-heavy — both lead on in-image typography, and [pro] accepts JSON-structured prompts that specify exact text positions, fonts, and styling for precise design work. Wrap the exact words in quotes, name the font and weight, give placement, and keep strings short for the cleanest lettering.
How many reference images can Flux use in one prompt?
FLUX.2 combines up to 10 reference images into a single output — for example a product shot, a face, and a style reference at once. This is how you keep characters and brand elements consistent across a set. Describe in the prompt what each reference contributes so the model knows how to blend them.
Does FLUX.2 support negative prompts?
No. FLUX.2 does not support negative prompts. Instead of listing what you don't want, describe what you do want — if you want a clean background, write "clean seamless white background" rather than "no clutter." Positive, concrete description is how the model was trained to work.
Do hex color codes work in Flux prompts?
Yes. Put a hex code such as #FF6B35 directly in the prompt and FLUX.2 reproduces that exact color, which makes it ideal for brand palettes. Name where the color goes — background, garment, or accent — and pair it with the color name so the intent is unambiguous.
Do I have to use every part of the formula?
No. Subject and action are the core; style and context are what turn an okay image into a great one. Start with subject plus action, then add style and context and one modifier at a time from the tables below, regenerating so you can see what each change does. Aim for the 30–80 word sweet spot.