How-To Guide
Flux AI
20 Examples
Flux AI Prompts: Complete How-To Guide + 20 Copy-Paste Examples
This is the how-to guide for writing Flux AI prompts — the 4-part SSST framework, 20 ready-to-use prompts organized by category, a Flux vs Midjourney prompt comparison, and the seven most common mistakes that kill Flux output quality. If you want curated examples and style references, see our Flux Prompts examples page. This guide is about how to build prompts from scratch.
Updated May 5, 2026
1,150 words
14 min read
What Makes Flux AI Different
Flux (developed by Black Forest Labs) is a rectified flow transformer — a fundamentally different architecture from the UNet-based models that powered Stable Diffusion and its derivatives. That difference in architecture has a direct consequence for how you write prompts.
Midjourney trained an aesthetic model that learns to interpolate between styles and apply its own aesthetic preferences to vague inputs. Flux was designed around literal language processing: it reads your prompt sequentially, weighting elements proportionally to how they appear in natural language. This means two things for how you write prompts:
- Order matters. Elements described first receive more compositional weight. Put your primary subject first, background elements last.
- Sentences outperform keywords. Comma-separated keyword lists lose relational information. "A woman standing in front of a red wall" conveys spatial relationship; "woman, red wall, standing" does not. Flux reads the former correctly and treats the latter ambiguously.
- No flag syntax. Flags like
--v6 --ar 16:9 --stylize 750 are Midjourney syntax. Flux ignores them as literal text. Aspect ratio and quality settings are configured in your Flux runner (Fal, ComfyUI, Replicate), not in the prompt text.
- Fewer negative prompts needed. Flux 1.1's architecture produces significantly fewer default artifacts than SDXL, so negative prompting is situational rather than essential.
The Flux Prompt Framework: SSST
Every strong Flux AI prompt follows a 4-part structure. Once internalized, you can construct any image prompt in under two minutes.
S
Subject
Who or what is the primary focus. Be specific about age, appearance, clothing, and expression for people. For objects, describe material, condition, and size.
A 30-year-old woman in a tailored navy suit, dark hair pinned up, confident expression
S
Scene
Where the subject is, what is happening, and how the environment looks. Include lighting conditions, time of day, weather, and relevant background elements.
standing in a glass-walled corner office at dusk, city lights visible through the windows behind her
S
Style
The visual style, medium, or aesthetic. Reference real-world equivalents: photography genres, painting movements, specific film stock types, or named artistic styles.
editorial fashion photography, shot for a business magazine, high-contrast lighting, desaturated palette
T
Technical
Camera and lens specifications, resolution descriptors, and rendering quality cues. These signal to Flux the level of technical precision expected.
Sony A7R V, 85mm f/1.4, shallow depth of field, 8K, ultra-sharp, no grain
Assembled: "A 30-year-old woman in a tailored navy suit, dark hair pinned up, confident expression, standing in a glass-walled corner office at dusk, city lights visible through the windows behind her, editorial fashion photography, shot for a business magazine, high-contrast lighting, desaturated palette, Sony A7R V, 85mm f/1.4, shallow depth of field, 8K, ultra-sharp." That is a complete, effective Flux AI prompt.
20 Copy-Paste Flux AI Prompts by Category
Each prompt below is structured using the SSST framework. They are ready to use as-is, or modify the Subject section to adapt to your specific use case.
Portraits
A 45-year-old man with silver hair and a weathered face, wearing a simple white linen shirt, looking directly into the camera with a calm, dignified expression. Plain charcoal gray studio backdrop, soft diffused Rembrandt lighting from the left. Documentary portrait photography, reminiscent of Steve McCurry. Phase One XF camera, 110mm f/2.8, medium format resolution, ultra-sharp eyes, slight grain for texture.
Rembrandt lighting creates a triangle of light under one eye — reference it by name for Flux to apply it precisely.
A young woman with curly auburn hair, freckles, wearing a cream-colored knit sweater, seated by a large window with soft morning light falling across her face from the right. She looks slightly off-camera with a contemplative expression. Lifestyle editorial photography, warm tones, golden hour color grade. Canon EOS R5, 50mm f/1.2, very shallow depth of field, bokeh background, 8K.
Specifying the direction of light (from the right) gives Flux explicit spatial information that keyword lists like "natural light, soft, warm" do not.
A bearded man in his 50s wearing a dark wool coat, standing in heavy rain on an empty city street at night. Neon reflections from a red sign illuminate the left side of his face while the right side remains in shadow. Film noir portrait, cinematic color grade with deep teal and orange tones. Leica M11, 35mm f/1.4, available light, rain texture on lens, high detail.
Describing the light source (neon from a red sign) gives Flux the information to render both the light color and the shadow shape realistically.
A professional woman in her 40s with sharp features and dark hair pulled back, wearing a structured black blazer over a silk white blouse. She stands against a seamless white background with clean even studio lighting, no shadows visible. Corporate headshot photography, high-key lighting, bright and clean aesthetic. Sony A9 III, 85mm f/2.8, 8K, crisp detail, perfect skin texture.
Adding "no shadows visible" removes Flux's tendency to add environmental shadows in portrait setups.
Architecture & Interiors
A minimalist living room with concrete floors, floor-to-ceiling windows overlooking a pine forest, a single curved white sofa, a low walnut coffee table with a single ceramic vase, warm afternoon sunlight casting long shadows across the floor. Architectural interior photography, Scandinavian modernism, neutral tones with natural wood accents. Hasselblad H6D, 24mm tilt-shift lens, perfectly straight verticals, ultra-high resolution.
Tilt-shift lens reference signals to Flux that vertical lines should be corrected — a key tell of professional architectural photography.
The exterior facade of a Gothic cathedral at blue hour, stone gargoyles visible along the roofline, stained glass windows glowing amber from within, a light mist rising from wet cobblestones in the foreground. Long-exposure architectural photography, dramatic sky with deep blue and purple tones, the cathedral lit by warm floodlights. Nikon Z9, 16mm f/2.8, 30-second exposure, sharp stone texture, star-burst effect on light sources.
Blue hour (20 minutes after sunset) gives Flux a precise time reference that produces a specific sky-to-building light balance.
A converted industrial loft apartment with exposed brick walls, polished cement floors, dark steel-frame windows, pendant Edison bulb lighting overhead, and a long rustic oak dining table set for six. Evening interior lighting, warm tungsten glow, urban industrial aesthetic mixed with warm hygge styling. Interior design magazine photography. Canon EOS R, 17mm f/4 tilt-shift, rich shadows, sharp throughout.
Mentioning "warm tungsten glow" versus "warm light" gives Flux a color temperature reference that produces a more specific, accurate rendering.
A futuristic glass and steel skyscraper with a curving parametric facade, photographed from street level looking up, surrounded by mature oak trees that contrast with the modern architecture. Shot on a clear day with a deep blue sky, reflections of clouds visible in the glass panels. Architectural photography, upward perspective, high dynamic range. Sony FE 14mm f/1.8, ultra-wide, 8K, no barrel distortion correction needed.
Including surrounding context (oak trees) gives scale and a designed contrast element rather than leaving the composition sterile.
Product Photography
A stainless steel luxury watch with a blue dial and sapphire crystal, resting on a folded dark navy fabric swatch, photographed flat from directly above. One edge is softly lit from a white light box on the right side, creating a gentle specular highlight along the case edge. Commercial product photography, clean minimal aesthetic, dark surface with subtle texture. Macro lens, 1:1 ratio, perfectly sharp across the entire face, no reflections on crystal.
"No reflections on crystal" prevents the common Flux artifact of a blown-out sapphire glass element in watch photography.
Three amber glass dropper bottles of face serum arranged on a white marble surface, surrounded by fresh sprigs of rosemary, a slice of lemon, and small white pebbles. Shot in bright natural daylight from a high 45-degree angle. Clean beauty product photography, spa aesthetic, soft shadows, fresh and natural color palette, no harsh specular highlights. Sony A7 IV, 90mm macro, f/8, 8K, crisp label text.
Specifying "crisp label text" helps produce legible bottle labels, which Flux occasionally blurs without an explicit instruction.
A kraft paper coffee bag with a black matte logo stamp, sitting on a wooden counter next to a small espresso cup and scattered whole coffee beans. Shot from a 30-degree angle in warm morning light from a nearby window. Artisan food and beverage product photography, warm amber tones, rustic coffeehouse aesthetic, shallow depth of field on the bag with beans softly blurred. Leica SL2, 75mm f/2, natural light, slight film grain.
Specifying the exact shooting angle (30 degrees) prevents Flux from defaulting to a straight-on shot, which loses the dimensional character of packaging.
A pair of matte black wireless headphones floating against a pure black background, illuminated by a single rim light from behind creating a subtle silver edge glow along the top arc of the headband. Technical product advertising photography, dark and premium aesthetic, perfect product symmetry, no dust or blemishes visible. Phase One XT, 110mm f/2.8, focus-stacked composite, every detail sharp, retouched commercial quality.
Rim lighting against black is a classic tech product photography technique. Flux replicates it accurately when described with "single rim light from behind."
Landscape
A dramatic mountain range at sunrise, the highest peaks cutting through a sea of low-lying clouds, golden light catching the snow-dusted upper ridgelines while the valleys below remain shrouded in purple-gray shadow. No human structures visible. Landscape photography, sweeping vista, shot from an elevated position looking east, rich color depth, high dynamic range. Nikon Z8, 24-70mm f/2.8 at 35mm, tripod, 5-exposure bracket blend, 8K.
Specifying "no human structures visible" is more effective than trying to describe a remote location abstractly — it gives Flux a clear exclusion rule.
Rolling sand dunes in a vast Saharan desert at dusk, warm amber and deep shadow alternating across the curved ridgelines, a single crescent moon visible in the darkening blue sky above the horizon. No vegetation, no footprints in the foreground. Fine art landscape photography, minimalist composition, strong graphic light-and-shadow contrast, David Muench style. Canon EOS R5, 400mm f/5.6, compressed perspective, extreme foreground-to-background sharpness.
Telephoto compression (400mm reference) flattens the dunes into a more graphic, less literal composition — Flux replicates this perspective correctly.
Dramatic basalt sea cliffs rising 80 meters above churning North Atlantic ocean waves that crash and foam against the rock base, shot from another cliff outcrop at eye level. Stormy overcast sky with low gray cloud, no blue sky visible. Moody landscape photography, high dynamic range, dark and powerful ocean atmosphere, Iceland coastal aesthetic. Sony A1, 16-35mm f/2.8 at 20mm, 1/500s to freeze wave motion, circular polarizer, 8K.
Including "1/500s to freeze wave motion" signals Flux to render sharp, detailed wave splash rather than motion-blurred water.
A narrow dirt path winding through a dense deciduous forest in peak autumn, maple and oak trees forming a canopy overhead with leaves in deep orange, red, and yellow, dappled morning sunlight filtering through to illuminate patches of the leaf-covered ground. No people visible. Fine art nature photography, warm autumnal palette, intimate forest atmosphere. Fujifilm GFX 100S, 45mm f/4, deep focus, rich medium-format rendering, slight film-like tonal curve.
Medium format camera references (GFX) signal a particular tonal richness and 3D rendering depth distinct from 35mm aesthetics.
Abstract & Conceptual
An abstract visualization of data flowing through a neural network, rendered as glowing blue-white light streams moving through a three-dimensional lattice of nodes and connections against a pure black background. The streams pulse with varying brightness suggesting data density. Digital art, scientific visualization aesthetic, ultra-clean render with no film grain. Octane render quality, 8K, volumetric light scattering on the streams, perfect mathematical symmetry in the node grid.
Referencing a specific render engine (Octane) signals Flux to apply physically accurate light behavior — particularly important for volumetric and glow effects.
Deep violet ink slowly dispersing into perfectly still clear water, captured at 1/8000 second to freeze the fractal tendrils of pigment mid-diffusion. Shot from directly above on a white seamless surface. Macro high-speed photography, scientific beauty aesthetic, the ink forming organic branching cloud patterns, no turbulence visible in the surrounding water. Canon MP-E 65mm macro, 5:1 magnification, high-speed flash, ultra-sharp ink edges, no motion blur.
The "1/8000 second" shutter reference tells Flux to render the ink with frozen clarity rather than the soft blur it defaults to without an explicit speed constraint.
A complex geometric sculpture made entirely of intersecting golden wireframe tetrahedra, suspended in a dark gallery space, illuminated by a single overhead spotlight that casts intricate shadow lattices on the white floor below. The sculpture appears to be approximately 1.5 meters tall. Fine art photography of a physical installation, minimalist gallery aesthetic, Olafur Eliasson exhibition style. Leica M11, 28mm f/2.8, long exposure, perfect shadow geometry on the floor.
Referencing a specific artist (Eliasson) gives Flux a clear aesthetic reference point for the installation style — more effective than abstract descriptions of that aesthetic.
A composite long-exposure photograph showing star trails as concentric arcs of white light around the north celestial pole, above a silhouetted rocky desert landscape. The foreground rocks are illuminated by warm amber light painting from a hidden lamp below the frame. Astrophotography, 4-hour equivalent star trail composite, the circular arcs of stars centered precisely in the upper third of the frame, deep black sky with no light pollution. Nikon Z6 III, 14mm f/1.8, ISO 800, stacked composite, no noise.
Specifying the composition position of the north pole ("upper third") gives Flux a precise compositional anchor — without it, the rotation center often appears centered, which looks less interesting.
Flux vs Midjourney: Prompt Style Comparison
| Prompt Element |
Flux AI Approach |
Midjourney Approach |
| Structure |
Full descriptive sentences with natural language syntax |
Comma-separated keywords, often noun-first |
| Subject order |
Primary subject first, modifiers follow in logical sequence |
Keywords ranked by weight; order matters less |
| Style reference |
Literal technical equivalents: "Rembrandt lighting from the left at 45 degrees" |
Aesthetic shorthand: "dramatic lighting, chiaroscuro" |
| Camera specs |
Embedded in prompt as descriptive context |
Often appended as keyword: "DSLR, 85mm, bokeh" |
| Flag syntax |
Not supported — configure in runner settings |
Essential: --ar, --stylize, --v6, --chaos |
| Negative prompts |
Situational, not required by default |
Often essential to suppress default artifacts |
| Optimal length |
40-80 words for most subjects |
20-50 words; shorter often works better |
| Realism ceiling |
Photorealistic — often indistinguishable from photography |
Stylized — usually readable as AI-generated |
Common Flux Prompt Mistakes and Fixes
Mistake 1: Keyword Lists Instead of Sentences
Wrong
woman, red dress, sunset, beach, bokeh, cinematic, 4k, beautiful
Fixed
A young woman in a flowing red dress standing at the shoreline of a beach at sunset, warm golden light behind her, shallow depth of field. Cinematic lifestyle photography. Sony A7 IV, 85mm f/1.4, 8K.
Mistake 2: Midjourney Flags in the Prompt
Wrong
mountain landscape at dawn --ar 16:9 --stylize 800 --v6 --quality 2
Fixed
A mountain landscape at dawn, golden light on the peaks, sweeping vista. [Set aspect ratio 16:9 in Fal/ComfyUI settings — not in the prompt text.]
Mistake 3: Subject Buried at the End
Wrong
Dark moody atmosphere, autumn leaves on the ground, forest path, cinematic light, aged man walking
Fixed
An aged man walking along a narrow forest path covered in autumn leaves, surrounded by dark moody atmosphere, cinematic low light filtering through bare trees. Editorial photography.
Mistake 4: Vague Quality Descriptors Without Technical Anchors
Wrong
hyperrealistic, 8K, masterpiece, highly detailed, best quality, award-winning
Fixed
Shot on Hasselblad H6D, 80mm f/2.8, ISO 100, medium format rendering, tack-sharp focus across the subject plane.
Mistake 5: No Lighting Direction Specified
Wrong
A portrait of a man, dramatic lighting, chiaroscuro effect
Fixed
A portrait of a man lit by a single light source positioned at 45 degrees to the left and above, casting a deep shadow on the right side of his face. Chiaroscuro portrait photography.
Mistake 6: Contradictory Instructions
Wrong
minimalist composition with many interesting objects and lots of texture and detail everywhere
Fixed
Minimalist composition: a single ceramic bowl on a bare white surface, soft texture visible in the bowl's glaze, everything else empty space. Architectural still life photography.
Mistake 7: Requesting Both Shallow and Deep Depth of Field
Wrong
portrait with beautiful bokeh and everything in sharp focus f/1.4
Fixed
Portrait with the subject in sharp focus and the background rendered as soft out-of-focus bokeh. 85mm f/1.4, subject distance 2 meters, background 5 meters behind.
Master Prompting Across Every AI Model
The framework on this page is one skill out of dozens. PromptSharp trains you systematically across Claude, ChatGPT, Gemini, Grok, Perplexity — and every image model — with daily exercises that build real prompting ability.
Start Learning for Free →
Model-agnostic · Duolingo for prompts · Cancel anytime
Frequently Asked Questions
What are Flux AI prompts? +
Flux AI prompts are text descriptions used to generate images with Flux, the open-weights image generation model developed by Black Forest Labs. Unlike Midjourney's flag-and-keyword syntax, Flux processes prompts as natural language sentences, reading them literally and sequentially. This means Flux AI prompts should be written as complete, descriptive sentences rather than comma-separated keywords, and every word in the prompt influences the output proportionally to how much visual weight the described element should carry in the image.
How long should a Flux AI prompt be? +
Flux AI prompts work best at 40-80 words for most subjects. Shorter than 40 words tends to leave the model underdirected — it fills in details arbitrarily, often producing inconsistent results. Longer than 100 words can cause Flux to dilute attention across too many elements, reducing the sharpness of the primary subject. The SSST framework (Subject, Scene, Style, Technical) naturally produces prompts in the 50-70 word range, which is the practical sweet spot.
Does Flux need negative prompts? +
Flux 1.1 and later versions largely do not need negative prompts to produce clean, high-quality outputs — unlike Stable Diffusion XL, where negative prompting was essential to suppress artifacts, bad anatomy, and watermarks. Flux's architecture produces fewer default artifacts. However, negative prompts can still be useful in specific situations: preventing certain compositional elements, suppressing specific color casts, or removing common default behaviors like vignetting in portrait shots. Use them when you have a specific problem to prevent, not as default boilerplate.
How is Flux different from Midjourney for prompting? +
The core difference is in how they parse language. Midjourney was trained with a strong aesthetic model that interpolates between styles and interprets vague prompts through its own aesthetic preferences. Flux processes prompts more literally, following the exact order and relative emphasis of words in your prompt. Flux rewards complete sentence structure, precise descriptive language, and explicit technical specifications. Midjourney rewards brevity and aesthetic shorthand; Flux rewards precision and completeness.
Can I use Flux AI prompts from Midjourney directly? +
Midjourney prompts will produce results in Flux, but usually not optimal results. Comma-separated keywords lose their relative weighting in Flux's sequential parser. Aesthetic shorthand like "cinematic" or "masterpiece" should be embedded in context rather than appended as standalone descriptors. Midjourney flag syntax is irrelevant to Flux. Translating a Midjourney prompt to Flux means expanding it into full descriptive sentences and replacing aesthetic keywords with their literal technical equivalents.
What is the best platform to run Flux AI prompts? +
Flux AI prompts can be run on several platforms in 2026. For free or low-cost access: Fal.ai (Flux Schnell for fast drafts, Flux Dev for quality), Replicate (pay-per-generation), and Krea.ai (free tier with daily limits). For local inference: Flux runs on Apple Silicon M2 Pro or better, or any CUDA GPU with 12GB+ VRAM using ComfyUI. For best quality without local hardware: Fal.ai running Flux 1.1 Pro produces the highest quality outputs and supports ControlNet, LoRA fine-tunes, and IP-Adapter for reference-guided generation.