AI Prompt Command Encyclopedia

A comprehensive handbook for mastering Midjourney, Stable Diffusion, Leonardo, Runway, Kaiber & more.

I. Core Parameters

These are the universal switches that control image structure, randomness, quality, and style. They work in nearly every major diffusion model and behave like camera or render settings.

Parameter	Description	Typical Values / Notes
`--ar`	Aspect ratio — defines the frame's shape.	`1:1` (square), `16:9` (landscape), `9:16` (portrait)
`--v`	Model version — selects algorithm generation.	Use latest for best lighting and anatomy.
`--q`	Quality level — how much GPU time is spent.	`0.25–2`; higher = sharper but slower.
`--s`	Stylize — how strongly the AI applies its own art sense.	`0–1000`; low = realistic, high = artistic.
`--chaos`	Variation/randomness in composition.	`0–100`; high = unexpected results.
`--seed`	Locks random generation for reproducibility.	Any integer; same seed = same structure.
`--iw`	Image weight — how much a reference image dominates text.	`0.1–2.0`; higher = follow image more strictly.
`--no`	Exclusion list — removes unwanted items.	`--no text, --no blur` etc.
`--style`	Model preset look.	`raw` for photorealism, others for thematic modes.
`--tile`	Creates seamless textures for pattern work.	Useful for fabrics and wallpapers.
`--stop`	Ends render early (0–100%).	50% yields soft unfinished feel.
`--uplight / --upbeta`	Different upscaling engines.	Light = smooth, Beta = detailed.
`--weird`	Experimental randomness knob.	`0–3000`; adds surreal artifacts.
`--cref`	Composition reference — anchors framing from an image.	Ideal for storyboards or consistent shots.
`::`	Prompt weighting — adjusts importance of elements.	`portrait::2 sunlight::1 fog::0.5`

Summary: Core parameters define how the AI interprets and structures your scene. Set aspect ratio and quality first, style and chaos second, then seed for repeatability.

II. Camera & Photography Controls

These terms simulate how a camera sees the world. Use them to control perspective, focus, light, and emotional tone just like a cinematographer would.

Lens Type and Focal Length

Lens	Purpose	Visual Effect
35mm	Natural field of view	Mild distortion, wide for streets or mid-shots
50mm	Human-eye equivalent	Neutral, documentary realism
85–100mm	Portrait lens	Flattering compression, blurred background
200mm+	Telephoto	Tight, cinematic framing

Rule: For people → 85mm; for architecture → 24–35mm.

Aperture & Depth of Field (f-stop)

f/1.4 → very shallow, background blur (bokeh)
f/11 → deep focus, everything sharp
Use "shallow depth of field" for portraits, "deep focus" for landscapes.

Camera Angles & Movement

Term	Meaning	Effect
Eye-level shot	Camera aligned with subject's eyes	Neutral realism
Low-angle	Camera looks up	Power, dominance
High-angle	Camera looks down	Vulnerability
Dutch angle	Tilted frame	Tension, drama
Close-up / Extreme close-up	Tight crop	Emotion, texture
Wide shot / Establishing	Shows context	Scene-setting

Lighting Language

Keyword	Definition	Used For
Soft light	Diffused source, smooth shadows	Portraits
Hard light	Strong shadows	Gritty realism
Backlit / Rim light	Light from behind subject	Cinematic outlines
Golden hour	Warm sunset tones	Emotional warmth
Neon lighting	Artificial colored light	Cyberpunk feel
Volumetric light	Visible beams through fog	Atmospheric depth

Composition Tools

Rule of thirds – subject off-center for tension.
Leading lines – guide viewer's eye.
Symmetry – balance and calm.
Negative space – minimalism.
Foreground elements – create depth.

Color & Mood

Cue	Mood Effect
Warm tones	Comfort, nostalgia
Cool tones	Isolation, calm
High contrast	Drama, intensity
Desaturated	Melancholy

Summary: Lens, light, and perspective define realism. The more precisely you mimic real photographic language, the less "AI" your image feels.

III. Loop & Motion Controls

These commands and cues are used for seamless loops, camera moves, and storytelling across frames. Perfect for AI video tools like Runway, Pika, Kaiber, or Stable Video Diffusion.

Loop Prompts

loop cue: — tells the AI to create a seamlessly repeating motion (example: waving flag, blinking lights, ocean waves).
seamless loop, infinite motion, perfect cycle — natural language variations for the same purpose.
beginning and end frame match — used for cinematic loops, avoids "jump cut."

Camera Motion Cues

Keyword	Meaning	Used For
camera pan	Horizontal camera sweep	Side motion across a subject
camera tilt	Vertical up/down motion	Reveal tall subjects
camera orbit	Circle around the subject	Dynamic 3D movement
dolly in/out	Camera moves closer or away	Emotional or spatial emphasis
tracking shot	Camera follows movement	Action scenes

Frame Timing Keywords

slow motion — emphasizes fluidity and emotion.
long exposure — adds motion blur trails.
hyperlapse / timelapse — accelerates perceived time.
frame-by-frame motion — used for animation consistency.

Loop-Optimization Parameters (for animated AIs)

Parameter	Purpose	Typical Value
`--loop`	Enables frame continuity	`true` / `1`
`--fps`	Frames per second	12–30 fps typical
`--motion`	Amount of camera/subject movement	`0.1–1.0`
`--stabilize`	Keeps horizon fixed	`on`
`--temporal`	Temporal coherence strength	`0.3–0.9`

Summary: Loop and motion commands let AI think in time, not just space. Use them when your goal is fluid storytelling — smooth repetition, consistent direction, and natural cinematic rhythm.

IV. Rendering & Material Systems

Rendering keywords control how materials, surfaces, and light behave. They imitate the physics of real-world cameras, CGI pipelines, and cinematography.

Render Engine Styles

Keyword	Description	Visual Feel
Octane Render	High dynamic range and cinematic contrast	Realistic, warm tone depth
Unreal Engine	Game-engine realism, perfect for cinematic video scenes	3D, hyper-detailed, interactive light
Arnold Render	Film production rendering standard	Balanced realism with natural shadows
Redshift Render	GPU-optimized cinematic lighting	Soft diffusion, subtle bloom
V-Ray Render	Architectural precision, clean lighting	Perfect for interiors, design objects
Blender Cycles / Eevee	Physical renderer vs real-time engine	Cycles = realistic; Eevee = fast previews

Lighting Techniques

Keyword	Effect	Usage
Global illumination	Simulates real light bounce	True-to-life indoor scenes
Ray tracing	Calculates physical light paths	Sharper reflections and refractions
Volumetric lighting	Visible beams of light	Fog, smoke, god-rays
Subsurface scattering	Light diffusion through skin	Realistic portraits, organic objects
Caustics	Light bending through glass or water	Used in ocean or bottle shots

Surface & Texture Keywords

Term	Description	When to Use
PBR (Physically Based Rendering)	Real material light response	Product renders, architecture
Displacement / Bump map	Microscopic surface detail	Stone, fabric, leather
Roughness map	Defines gloss vs matte areas	Metal, skin, plastics
Specular reflection	Shiny surface highlight control	Polished products
Ambient occlusion	Contact shadow realism	Fills dark crevices naturally

Post-Processing & Film Simulation

Film grain: Adds analog texture; prevents sterile look.
Chromatic aberration: Subtle lens color shift at edges.
Vignette: Darkens frame corners, cinematic focus.
Bloom: Soft light glow from bright highlights.
Tone mapping: HDR compression for natural exposure.

Summary: Rendering terms tell the AI *how* to visualize light and surface, not just *what* to show. Combine one render engine, one lighting model, and a few post effects for cohesive realism.

V. Artistic Styles & Mediums

These keywords decide how the AI interprets form, texture, and visual tone. They simulate traditional art styles, genres, and movements to shape emotional identity and aesthetic intent.

Medium & Technique Keywords

Keyword	Imitates	Visual Feel
digital painting	Stylized digital brushwork	Soft, painterly, concept-art style
oil painting	Classic canvas oil texture	Rich color depth, expressive brush strokes
watercolor	Translucent color washes	Dreamlike, fluid edges
charcoal sketch	Rough black shading	Dramatic contrast, sketch aesthetic
ink drawing	Precise line work	Graphic novel / manga vibe
photographic realism	Camera-grade realism	Photo-level detail

Art Movements

Style	Description	Mood / Example Use
Baroque	High contrast, ornate composition	Dramatic and emotional
Impressionism	Soft light, visible brush texture	Dreamy nostalgia
Noir	Monochrome with strong shadow contrast	Mystery, detective themes
Cyberpunk	Neon, futuristic urban decay	Technological dystopia
Minimalist	Few elements, negative space	Elegance and calm
Surrealist	Dreamlike distortions	Fantasy or conceptual art
Retro / Vintage	Old color palettes and textures	1950s–80s nostalgia
Monochrome / Sepia	Single-color tone	Historic or emotional restraint

Color & Lighting Emotion

Color Style	Emotional Effect
Muted palette	Quiet, melancholic
Vibrant tones	Energy and optimism
Pastel colors	Soft, romantic, nostalgic
Desaturated / Bleached	Gritty realism
High contrast	Drama and intensity
Low contrast	Dreamlike and gentle

Genre Tags

Editorial: Fashion or magazine aesthetic, clean and composed.
Cinematic: Film-like lighting and aspect ratio, storytelling tone.
Documentary: Realistic, candid lighting, and natural poses.
Fantasy: Myths, magic, surreal color palette.
Science fiction: Futuristic environments, glowing technology.
Dark aesthetic: Shadow-heavy, mysterious, dramatic mood.

Summary: Art style words control visual language. Limit yourself to one or two dominant styles per prompt — mixing too many (e.g. "cyberpunk watercolor oil photo") confuses the model and weakens focus.

VI. Advanced & Hidden Controls

These parameters appear in pro-level interfaces (Stable Diffusion, ComfyUI, Invoke, Leonardo, etc.). They control the math behind generation — diffusion steps, randomness, and structure references.

Diffusion Process Parameters

Parameter	Purpose	Typical Range / Value
`CFG scale`	"Classifier-free guidance" — how strictly AI follows text.	6–10 for balanced realism; lower = creative freedom.
`Steps`	Number of diffusion iterations.	20–60 for photos, 80+ for detailed art.
`Sampler`	Noise-solving algorithm (defines detail flow).	`DPM++ 2M` or `Euler a` are most reliable.
`Clip skip`	Skips layers of the text encoder.	1–3; higher gives looser interpretation.
`Eta (η)`	Controls randomness strength.	0–1; lower = stable, higher = creative variation.

Image-to-Image / Refinement Settings

Parameter	Purpose	Range
`Denoising strength`	How far new image diverges from input.	0.2–0.8 typical; 1.0 = full new render.
`VAE`	Color-decoding model.	Match to model for correct tone; wrong VAE = washed colors.
`Refiner`	Post-process model for extra sharpness or realism.	Optional; use when base lacks detail.

Control Systems

System	Function	Example Use
ControlNet	Transfers structure from pose, depth, or scribble maps.	Exact body posture, architecture lines, composition.
IPAdapter	Style or face reference matching without ControlNet.	Maintains likeness or brand identity.
LoRA	"Low-Rank Adapter" mini-model that injects styles or characters.	0.5–1.0 strength for natural integration.
Textual Inversion	Custom token learned from a dataset.	One word = entire visual concept.

Stability Tools

Seed Lock: Reuse seed across frames for consistent faces and lighting.
Batch Size: Multiple variations per prompt (useful for A/B testing).
Restore Faces: AI fix for distorted features (may over-smooth).
Highres Fix: Two-pass upscale for large, clean images.

Summary: Advanced controls are where creativity becomes engineering. Keep CFG moderate, sampler stable, and LoRA weights light — the best results come from balance, not brute force.

VII. Practical Prompting Examples

Each example below shows a functioning prompt, followed by analysis and recommendations for improvement.

Example 1 — Cinematic Western Portrait

A rugged outlaw standing on a dusty desert road at sunset, cinematic lighting, 85mm lens, film grain, Octane render --ar 21:9 --v 6 --q 2 --s 120

Why it works: Structured clearly — subject first, environment second, technical third.
Camera and lighting cues (85mm, sunset, cinematic) anchor realism.
Render engine (Octane) and film grain create a believable physical world.
Stylize 120 gives artistic warmth without distorting anatomy.

Tip: Reduce --s below 80 for a more documentary feel; increase to 250 for artistic drama.

Example 2 — Cyberpunk Street Scene

Rain-soaked city alley, neon reflections, lone figure with umbrella, cinematic perspective, Unreal Engine lighting --ar 16:9 --q 2 --chaos 15 --v 6

Why it works: Focused composition — atmosphere first ("rain-soaked"), then subject, then lighting.
Unreal Engine defines 3D realism and dynamic reflections.
Chaos 15 introduces minor variations in lighting for life-like unpredictability.
Aspect 16:9 fits film context; --q 2 keeps clarity in low-light scenes.

Example 3 — Product Aesthetics (Studio)

Matte black headphones on reflective marble surface, softbox lighting, minimalist product photography, 50mm lens --ar 3:2 --v 6 --style raw --q 1.5

Why it works: Balanced lighting and environment — clean studio tone.
Softbox and 50mm mimic real commercial photography setup.
"Raw" style disables over-artistic filters for accurate materials.
Marble surface adds subtle class and texture without clutter.

Improvement: Add "reflections softened by fog" for luxury mood; lower contrast for elegance.

Example 4 — Fantasy Concept Art

Ancient warrior queen on mountain peak, storm clouds swirling, golden armor glowing in lightning, oil painting style --ar 2:3 --v 6 --s 300 --q 1.5

Why it works: Powerful visual verbs ("glowing," "swirling") add energy.
Oil painting + stylize 300 merge painterly texture with epic scale.
Vertical 2:3 ratio complements tall composition (figure + storm sky).

Tip: Use --chaos 25 to explore different storm formations and lighting angles.

Example 5 — Realistic Portrait (Studio)

Close-up portrait of woman in soft window light, shallow depth of field, natural skin texture, f/1.8, realistic color tones --ar 4:5 --v 6 --style raw --q 2

Why it works: Uses precise photographic cues (f/1.8, window light, 4:5 aspect).
Natural lighting and raw style make AI output indistinguishable from DSLR photography.
Shallow focus and color tone cues add human warmth.

Recommendation: Add --seed to lock face geometry across variations for consistent character identity.

Summary: Always write prompts like you're describing a film scene to a crew — who, where, lighting, mood, and camera setup. Simplicity and clarity beat complexity every time.

VIII. Quick Reference Table

This summary condenses all essential prompt controls, grouped by function. Use it as a daily reference for precision and efficiency.

Category	Parameter / Keyword	Function	Typical Range / Note
Format	`--ar`	Aspect ratio	1:1, 16:9, 9:16
Format	`--v`	Model version	Use latest release
Quality	`--q`	Render quality	0.25–2
Style	`--s`	Stylization strength	0–1000
Variation	`--chaos`	Randomness	0–100
Seed	`--seed`	Reproducible randomness	Integer value
Reference	`--iw`	Image weight	0.1–2.0
Exclusion	`--no`	Exclude elements	--no blur, --no text
Tile	`--tile`	Seamless output	Pattern creation
Stop	`--stop`	End render early	10–90%
Upscale	`--uplight / --upbeta`	Upscaling mode	Light = soft; Beta = detailed
Experimental	`--weird`	Adds creative chaos	0–3000
Composition	`--cref`	Reference framing	Image-based layout
Weighting	`::`	Prompt emphasis	subject::2 lighting::1.2 fog::0.5

Camera & Motion

85mm lens – portrait compression
f/1.8 – shallow focus, blurred background
eye-level – neutral power balance
low-angle – strength and dominance
camera orbit / pan / tilt – dynamic cinematic motion
loop cue: – seamless animation loop

Rendering & Light

Octane / Unreal / Redshift: different render lighting engines
Global illumination: realistic bounce lighting
Volumetric light: beams, fog, atmosphere
Subsurface scattering: natural skin glow
Film grain / vignette: analog realism

Art Styles

oil painting, watercolor, charcoal, digital art
cyberpunk, noir, baroque, minimalism
vintage, monochrome, surrealism, cinematic

Advanced

CFG scale: text obedience (6–10 standard)
Steps: generation iterations (20–60 typical)
Sampler: noise solver (DPM++ 2M, Euler a)
Clip skip: looser concept control (1–3)
LoRA / ControlNet / IPAdapter: fine control systems
Denoising strength: image-to-image variation (0.2–0.8)

Summary: The Quick Reference Table is your cheat sheet. Always adjust --ar, --q, --s, and --chaos first — they define structure, clarity, and creative balance.

IX. Critical Prompting Insights

This final section covers the invisible logic of prompting — the psychology, structure, and hidden rules that separate casual users from true visual directors.

1. Prompt Hierarchy

AI reads prompts by concept clusters, not grammar. Always move from general to specific: subject → environment → lighting → camera → mood → technical. Group related descriptors and use weighting (::) to control emphasis.

2. Order of Operations

Prompts are parsed left to right by meaning weight. The first nouns influence composition most. Place critical subjects early, and technical tags (like --ar or --q) last.

3. Contrast Through Language

Replace vague terms ("cinematic," "dramatic") with contrast drivers: "soft light on rough metal," "dark figure in glowing fog." Realism lives in differences, not adjectives.

4. Dataset Bias

Models favor idealized Western imagery — smooth faces, centered symmetry. Counteract it with realism cues: "uneven lighting, asymmetrical pose, candid expression." That language breaks bias.

5. Prompt Density

Fewer words = clearer vision. Too many adjectives cause "semantic noise." Remove anything that doesn't change the mental picture.

6. Context Law

AI doesn't imagine, it averages. Always define setting words (fog, rain, dusk, studio) to anchor your scene in physical space. Without context, subjects float in voids.

7. Randomness Control

Parameters like chaos, eta, or CFG adjust exploration vs stability. High values = creative mutation; low = precision. Tune depending on experiment vs production.

8. Seed Discipline

Fix a seed for continuity in series or video work. Changing it changes identity. Always record your seed and metadata with good results.

9. Debugging Prompts

When results fail, isolate variables: remove lens or lighting cues first, then reintroduce one at a time. Treat prompts like film sets — one fix per test.

10. Metadata Logging

Save every great prompt with seed, sampler, CFG, and model name. If you can't reproduce it, you didn't really create it.

11. Iterative Prompting

Work in passes: base prompt → img2img → refined version. Like multiple exposure layers in film. Don't expect masterpiece in one step.

12. Ethics & Copyright

Learn technique, not imitation. Avoid replicating artists' names or signatures — study their lighting and composition instead. Mastery isn't mimicry.

13. Emotional Layer

Emotion keywords outperform technical ones: "tired," "restless," "hopeful" subtly guide pose and lighting. Every good frame carries feeling first, detail second.

14. Knowing When to Stop

AI can overwork images. Professionals stop once intent is captured — not perfection. Save early outputs; first drafts often hold strongest mood.

15. Think Like a Director

You're not writing text for a bot; you're directing a crew: camera (composition), lighting (mood), actors (subject), and set (environment). When that mental shift clicks, AI becomes your cinematographer — not your toy.

Summary: Prompting mastery is less about words and more about direction. You're shaping meaning, not syntax — structure, emotion, and light do 90% of the storytelling.

X. Visual Language Reference Library

This section is a high-density visual vocabulary. These keywords do not describe objects. They describe perception. Use them to steer light behavior, camera physics, composition logic, material realism, and emotional tone.

1. Lighting & Mood

Keyword	Effect	When to Use
golden hour	Warm low-angle sunlight	Emotion, nostalgia, cinematic warmth
blue hour	Cool twilight ambience	Calm, melancholy, urban scenes
soft diffused light	Gentle shadows	Portraits, beauty, realism
harsh midday sun	Strong contrast, hard shadows	Documentary, realism, tension
backlit / rim lighting	Subject edge glow	Separation, cinematic silhouette
volumetric light / god rays	Visible light beams	Atmosphere, depth, drama
neon-lit / candlelit / moonlit	Single-source mood lighting	Noir, fantasy, intimacy
high-key / low-key lighting	Bright vs shadow-dominant	Commercial vs dramatic tone
Rembrandt / butterfly / split	Classic portrait lighting setups	Professional facial sculpting
chiaroscuro	Extreme light-dark contrast	Painterly drama, baroque feel

2. Camera & Lens

Keyword	Meaning	Visual Outcome
shallow depth of field	Background blur	Subject isolation
deep focus	Everything sharp	Environmental storytelling
85mm portrait lens	Compression, flattering	Faces, fashion
35mm street lens	Wide realism	Documentary, lifestyle
macro 1:1	True close focus	Textures, insects, products
anamorphic lens flare	Cinematic streak highlights	Film look
long exposure / motion blur	Time compression	Traffic, water, motion
chromatic aberration / vignetting	Lens imperfection	Analog realism

3. Composition & Framing

Technique	Purpose	Psychological Effect
rule of thirds	Balanced tension	Natural viewing comfort
centered composition	Stability	Authority, calm
dutch angle	Tilted horizon	Unease, chaos
bird’s-eye / worm’s-eye	Extreme perspective	Power imbalance
close-up / extreme close-up	Detail focus	Emotion, intimacy
establishing shot	Scene context	Narrative clarity

4. Color & Palette

Scheme	Effect	Use Case
monochrome / black and white	Form over color	Timeless, graphic
pastel / muted tones	Soft emotional range	Nostalgia, calm
neon palette	High energy contrast	Cyberpunk, nightlife
teal and orange	Hollywood contrast	Cinematic storytelling
desaturated	Reduced color noise	Gritty realism

5. Materials & Textures

Material keywords define micro-surface behavior. They strongly affect realism when combined with proper lighting.

polished chrome, brushed aluminum, stainless steel
aged brass, oxidized copper patina
weathered wood, marble veining, terrazzo chips
frosted glass, stained glass
velvet nap, satin sheen, silk fabric, linen weave

6. Environment & Weather

Environmental cues anchor scenes in physical reality and prevent floating subjects.

foggy, misty, overcast, stormy sky
light drizzle, heavy rain, puddles and reflections
snow flurries, fresh snowfall, blizzard whiteout
dust storm, sandstorm, humid haze, smoggy air
aurora borealis, starry night sky

7. Art Style & Medium

Medium keywords override realism rules and define rendering logic.

oil painting impasto, watercolor wash, gouache poster style
ink line art, charcoal sketch, graphite pencil
digital matte painting, vector flat style
pixel art, low-poly 3D, voxel art
cel-shaded anime, manga screentone, comic halftone

8. Clothing & Fashion

Wardrobe descriptors subtly influence pose, silhouette, and attitude.

streetwear, haute couture, avant-garde fashion
business formal, smart casual, bohemian
gothic, punk, grunge, retro 70s, vintage 90s
fantasy armor, cyberpunk attire, steampunk outfit

9. Camera Models

Camera model names inject dataset-level bias for color science and contrast.

Leica M6, Hasselblad 500 C/M, Rolleiflex 2.8F
Pentax 67, Mamiya RZ67, Contax T2
Canon AE-1, Nikon F3, Polaroid SX-70
Fujifilm X100V, Canon 5D Mark II, ARRI Alexa

10. Analog Camera Effect

Analog descriptors add imperfection, memory, and emotional softness.

35mm film photograph, point-and-shoot flash
grainy film texture, slight color shift
low contrast look, soft shadows
nostalgic analog mood, Y2K aesthetic
shot on Kodak Portra 400

11. Film Stocks & Emulations

Film stock names primarily influence color curves and grain behavior.

Kodak Portra 160 / 400
Kodak Ektar 100, Gold 200
Kodak Tri-X 400, T-Max 400
Kodak Vision3 250D / 500T
CineStill 800T / 50D
Fujifilm Velvia 50, Provia 100F

Summary: This library is not meant to be stacked blindly. Choose one lighting logic, one lens logic, one color philosophy, and one material direction. Overloading keywords reduces clarity instead of increasing realism.