AI Prompting Guide

Tips

• Focus on Visual Continuity
  • Veo 3.1 Advanced connects moments across shots. Describe clear visual flow between scenes.
    Example: "A dancer twirls through a spotlight. The light fades and transitions into sunrise over a city rooftop where she finishes her pose in silhouette."
• Use Camera Transitions
  • Integrate cinematic camera terms and scene bridges:
    Cross-fade, dolly out, tilt up, match cut, fade-through-color, seamless zoom.
    Close-up, wide shot, slow motion, dolly in, aerial view, handheld cam, tracking shot.
    Example: "Camera dolly-out from a tight close-up of a painter’s brush, cross-fading into a wide aerial of the completed mural."
• Describe Sound & Emotion
  • Set tone with layered audio and emotion cues:
    Soft piano, urban ambience, thunderstorm, whispering wind, emotional crescendo.
    Example: "Rain pours as soft piano plays. A man says: I still remember her voice. Thunder rumbles, echoing his words."
• Show Realistic Motion
  • Veo 3.1 handles natural physics and fluid animation.
    Add verbs that create movement:
    Running, spinning, drifting, falling, lifting, looking back, turning, breathing.
    Example: "The camera follows behind as she runs through the forest. Leaves scatter underfoot and her breath fogs in the cold air."
• Add Style & Consistency
  • Choose your storytelling mood or art direction:
    Cinematic, photoreal, painterly, anime, retro, dreamlike, documentary.
    Example: "Dreamlike sequence — soft focus, glowing edges, ethereal lighting, lens flare drifts across the frame."
• Use Transitions (New in 3.1)
  • Now supports smooth multi-scene continuity and sound-matched transitions:
    Fade-to-black, motion-match dissolve, synchronized audio cross-fade.
    Example: "The roar of a train fades into ocean waves as the scene transitions from a subway tunnel to a coastal cliff."
• Keywords to Include
  • 4K, cinematic, volumetric light, shallow depth of field, natural shadows, dynamic motion, seamless transitions, audio sync, atmospheric, high consistency, soft focus, ambient score.
  • Image-to-Video Tips. If using a photo:
    Describe motion only, avoid restating static elements.
    For sequential frames, specify start-to-end motion.
    Good Example: "Start frame: portrait of a woman gazing out a window. End frame: camera slowly pulls back as sunlight fills the room, blending with soft background music."
• Include Audio & Dialogue
  • Add background music, ambient sounds, and speech for full realism.
    ✅Correct: He says: Time waits for no one
    🚫Avoid: He says: “Time waits for no one”
• Need inspiration? Just start with this format:
  • [Subject] + [Action] + [Scene] + [Camera] + [Mood] + [Audio] + [Transition]

Tips & Prompting Stylie

• Ultra-Realistic + Cinematic Physics + Emotional Atmosphere.
  Best For: Real-world logic, lifelike texture, natural motion, subtle detail, grounded physics, and documentary or lifestyle scenes.
• Motion Description: Always clearly describe what’s happening, e.g. “A dog runs across a sandy beach and splashes into the water.”
• Use Real-World Detail: Describe setting and physical textures (e.g. “Wet cobblestone street reflecting neon lights,” “wood grain,” “soft sunlight through dusty air.”)
• Scale & Physics: Be creative with physical ideas: “A coin spins in slow motion on a glass table,” “Rain falling onto hot metal sizzles into steam.”
• Emotion & Vibe: “Warm, nostalgic lighting,” “peaceful vibe,” “slow and romantic.” Mood is as important as subject.
• Camera Movement: Include “slow dolly out,” “macro close-up,” “overhead drone shot,” “push in.”

Tips & Best Practices

• Structure Your Prompt: Always include subject, subject description, subject movement, scene, scene description, and optionally—camera style, lighting, and atmosphere.
• Simple Action Language: e.g., “A knight walks through misty woods at dawn.” or “A woman walks through falling cherry blossoms.”
• Motion Control (Image2Video):, Combine subject + motion with background + motion. E.g., “A woman holds a cup of coffee, subtle steam rising, background of a rainy window.”
• Physics-Driven Events: e.g., “Raindrops splashing in slow motion”, “Book falling, pages flipping with wind.”
• 360 & Split-Screen: Use “360-degree spin”, “4 camera angles for 4 seasons”, or “multi-view split screen”
• Avoid Numbers: Use terms like “a group of birds” instead of “7 birds” for best results.
• Image2Video Tip: Describe motion explicitly, like “Mona Lisa slowly lifts sunglasses with her right hand.”

Tips

• Use structured prompts i.e. organize Text-only prompts into [camera movement]: [scene setup]. [details].
• All prompts should be direct and easily understood, not conceptual (e.g. "a man vigorously typing on the keyboard" instead of "a man hacking")
• Prompts should be descriptive, not conversational or command-based (e.g. "a dog playfully runs across the field" instead of "can you make a video of a dog playing in the field" nor "add dog to the image")
• Prompts should use positive phrasing (e.g. use "a clear sky" instead of "a sky with no clouds", or use "the camera remains still" instead of "no movement")

Tips

• Prompt formula: Subject + Subject Motion + Scene + Camera Movement + Aesthetic Atmosphere.
• Precise Camera Movement: Detailed instructions like "The camera descends slowly, wrapping around to focus on the subject's face."
• Atmospheric Aesthetics: Specify ambiance with terms like "natural color tones" or "neon reflections in a bustling city."
• Transformation: Describe environmental changes, e.g., "A foggy forest clearing with light beams piercing through."

Tips

• Lead with movement: Use action verbs like twirl, leap, glide, spin, wave, stomp.
• Set the mood: Emotion and lighting define the feeling of the performance.
• Direct the camera: Describe where the camera is, how it moves, and what it captures.
• Use real-world lighting or stylized environments: Spotlight, neon glow, sunset, fog, etc.
• Balance detail and flow: 1 action + 1 emotion + 1 visual + 1 camera move = powerful prompt.

Tips

• Add a descriptive text prompt and define the elements in your scene.
  • Subject + Subject Description + Action + Environment
• Set emotional atmosphere like the mood or tone and give a proper reference style like cinematic or cartoon

Tips

• Use this tool for good physical accuracy, high dynamics, and scene consistency.
• A compelling AI video prompt consists of 7 components:
  • 1. Subject - define your star: "A young woman with flowing red hair", "A sleek electric sports car", "A majestic eagle in flight".
  • 2. Scene - set your stage: "In a neon-lit cyberpunk cityscape", "Amidst a snow-covered forest at dawn", "Inside a minimalist modern apartment"
  • 3. Motion - bring life to your video: "Gracefully dancing through falling autumn leaves", "Rapidly accelerating along a coastal highway", "Smoothly transitioning between yoga poses"
  • 4. Camera movement - direct your shot: "Slow upward tilt revealing the cityscape", "Smooth tracking shot following the subject", "Dramatic circular pan around the scene"
  • 5. Atmosphere - create the right mood: "Mysterious and ethereal atmosphere", "Energetic and vibrant mood", "Calm and serene ambiance"
  • 6. Lighting - perfect your illumination: "Soft, warm sunlight filtering through trees", "Sharp, contrasting shadows from street lights", "Diffused, ethereal glow from above"
  • 7. Shot composition - frame your vision: "Close-up shot focusing on emotional expression", "Wide landscape shot emphasizing scale", "Low-angle shot creating dramatic perspective"
• Use Evocative Language: e.g. use "Sun-drenched" instead of "A bright day", use "Surging," "Gliding" instead of "Moving quickly".
• Paint with Details: e.g. use "Ancient redwoods towering into misty heights" instead of "A forest", use "A Victorian mansion with intricate wrought-iron details" instead of "A nice house"
• Maintain Visual Consistency: Keep lighting conditions unified throughout the scene, ensure color themes work together, match mood elements across all scene components, maintain realistic physics and movement patterns.
• Write Clear, Direct Descriptions: Focus on concrete visual elements, describe exactly what should be seen, structure descriptions from main subject to supporting details, use precise spatial and directional terms.

Tips

• Quite capable open-source tool from Alibaba with good understanding for physics, adherence to prompts, and stunning videos especially in the anime category.
• Be descriptive to set the characters’ appearance, the background settings, and the desired motions.

Tips

• Keep it short, vivid, and visual — imagine you’re describing a movie shot to someone holding the camera.
• Focus on who / where / what’s happening
• Use strong visual words — colors, lighting, vibe
• Add camera moves (close-up, wide, slow pan)
• Keep one clear action per shot
• Optional: Add emotion or cinematic style

Best Practices

• Think in Shots, Not One Block — Structure your prompt as individual shots with durations and descriptions that follow a narrative.
• Anchor Your Subjects — Define core subjects early with consistent descriptors so Kling maintains continuity across camera moves and different shots.
• Describe Motion Explicitly — Instead of vague verbs, describe how the camera moves and how the subject moves over time.
  • Slow tracking shot following subject from left to right
  • Medium shot, camera holds while subject pauses
  • Push-in from wide to close with subtle focus shift
• Use Native Audio Cues — When you include dialogue or specific soundlines, tag the speaker and tone.
  Example: [Lead, calm voice]: “Welcome to our Iftar.”
• Take Advantage of 15-Second Duration — Kling 3 can generate longer sequences with evolving actions. Describe progression over time for smoother flow and fewer artifacts.

Best Practices

• Clear Physical Actions — Pixverse responds best to concrete, physical verbs:
  • walking, lifting, pushing slowly, sipping, turning head slightly
  Avoid vague or emotional-only words like “beautiful” or “dramatic” without context.
• Camera Notes — Pixverse handles camera cues well if they’re concrete:
  • Slow push-in — gradual forward movement
  • Side tracking — camera slides left/right
  • Static hold then move — helps avoid floating feel
  • Natural shifts — small depth changes

Use with shallow depth instructions if you want specific focus: Shallow depth of field, subject in focus, background soft blur

Example Prompt: A modern cafe scene, barista gently pours latte into cup, steam rising only from the coffee, warm indoor lighting, slow handheld push-in camera from right, shallow depth of field, natural micro head movement, slight hand adjustments, realistic motion pacing, no synchronized multiple actions, ambient café background.

Tips & Prompting Style

• Lead with intentional motion: Use precise action verbs like spin, glide, strike, snap, flow, leap — avoid generic “dancing.”
• Define performance energy: Emotion should drive the movement — graceful, aggressive, playful, powerful, cinematic.
• Direct the camera clearly: Always describe position, movement, and framing (orbit, tracking, close-up, pull-back).
• Design the environment: Let surroundings react — dust, fabric, light, reflections, particles.
• Balance control and clarity: 1 subject + 1 core motion + 1 emotion + 1 environment + 1 camera move = strong output

Tips & Prompting Style

• Think in sequences: Write prompts like mini-scenes or multi-shot narratives, not single moments.
• Control continuity: Describe how scenes transition — cut, pan, reveal, or shift focus.
• Be production-aware: Specify duration, pacing, and resolution-style output (cinematic, ad, product shot).
• Use camera direction heavily: Wan 2.7 responds strongly to shot design and framing.
• Leverage realism: Add textures, reflections, physics, and environment interaction.
• Balance structure and flow: 1 subject + 1 action + 1 environment + 1 camera + 1 progression = strong output

Tips & Prompting Style

• Think like a director + editor: PixVerse V6 is built for multi-shot storytelling, not just single clips.
• Use camera or effect (not both): Choose between camera movement OR stylized effects, not both together.
• Leverage transitions: Use first frame → last frame to control scene evolution precisely.
• Use multi-clip mode: Create connected sequences with consistent characters and flow.
• Add audio always: Native audio + video generation in one pass is a major strength.
• Be cinematic with control: Include lens, depth, motion, and framing — PixVerse responds strongly to these.
• Balance structure and direction: 1 subject + 1 action + 1 camera + 1 environment + 1 transition = strong output

Tips & Prompting Style

• Think like short-form content: Grok Imagine is built for 6–10s viral clips, not long cinematic scenes.
• Use audio-driven prompts: Always consider sound, music, or dialogue — it generates audio automatically.
• Be expressive, not technical: Focus more on what happens + how it feels, less on complex structure.
• Leverage multi-modal inputs: Combine text + images + references for better control and consistency.
• Keep motion simple but impactful: One strong action works better than multiple complex actions.
• Balance creativity and clarity: 1 subject + 1 action + 1 vibe + 1 sound = strong output

Tips & Prompting Style

• Think short-form + fast: Designed for quick 1–10s clips, ideal for rapid iteration and social content.
• Prioritize clarity over complexity: Simple, well-structured prompts perform better than overloaded scenes.
• Use motion intensity: Control energy using subtle / medium / high movement (important for Vidu).
• Leverage native audio: Add music, ambience, or sound effects directly in prompt.
• Focus on smooth motion: Vidu excels at stable movement + clean camera flow.
• Balance speed and structure: 1 subject + 1 action + 1 environment + 1 camera + 1 audio = strong output

Tips & Prompting Style

• Think from the image outward: The input image is the foundation, your prompt should describe how it evolves.
• Control motion clearly: Define exact movement (walk, turn, rotate, explode, reveal).
• Use start → end logic: If using end frame, describe transition between frames.
• Leverage multi-shot prompts: Break scenes into multiple shots for better storytelling.
• Use elements for consistency: Reference characters/objects using @Element1, @Element2 for stable identity.
• Add audio when needed: Native audio + voice control is supported for richer output.
• Balance structure and control: 1 image + 1 motion + 1 camera + 1 progression = strong output

Tips & Prompting Style

• Think quick drafts: Veo 3.1 Lite is optimized for fast idea → video workflows, not heavy cinematic control.
• Keep prompts simple: It understands natural language, no need for complex structure or technical wording.
• Focus on short clips: Best for 4–8 second videos, ideal for reels, ads, and concepts.
• Use iteration mindset: Generate multiple versions and refine — it’s built for testing ideas quickly.
• Leverage image inputs: Start from an image to create better controlled motion outputs.
• Add sound intent: You can guide sound effects or vibe, even in simple prompts.
• Balance simplicity and clarity: 1 subject + 1 action + 1 vibe + 1 camera = strong output

Tips & Prompting Style

Brevity beats detail. HappyHorse works best with ~20 words: Subject, Action, Setting, and one camera or atmosphere cue. More words = worse motion quality (faces, hands, movement degrade).

Standard format: [Subject] [does action] in [setting], [time of day], [one camera or atmosphere cue].

Example: A young woman walks through a neon-lit street at night, wet reflections on asphalt, slow tracking shot.

Golden Rules

• Start with SUBJECT + ACTION — Always lead with what’s happening: “A man runs through a forest”, “A product rotates in mid-air”, “A girl turns and smiles”.
• Keep it SHORT (20–30 words max) — Short = clean motion. Long = broken physics. Overloading reduces realism dramatically.
• Use ONLY ONE strong cinematic cue — Pick one: camera (slow dolly, orbit, tracking), lighting (neon reflections, backlight), or lens (telephoto, shallow depth). Don’t stack multiple cinematic ideas.
• Use SPECIFIC visuals (not vague words)
  • ❌ Avoid: “beautiful”, “stunning”, “epic”, “ultra detailed”
  • ✅ Use: “neon pink reflections”, “warm amber backlight”, “wet asphalt”
• Camera motion = HIGH IMPACT — HappyHorse is especially strong at tracking shots, dolly-in, orbit shots, and aerial movement. Add camera at the END of prompt.
• Avoid “AI prompt habits” — Don’t use keyword lists, JSON, booru tags, or weighted prompts. Use simple natural English sentences only.

OLD: Create a cinematic video of a person walking in a city with dramatic lighting and high quality visuals.

NEW: A man walking in a modern city during daytime. The camera follows him from behind. Natural movement, realistic lighting.

Advanced Prompt Structures

• Short cinematic prompt (best default): A premium product bottle rotates in a dark studio, soft reflections on glossy surface, slow camera push-in.
• Camera-focused prompt (40–60 words): A woman in a red coat walks through a rainy city street at night. The camera tracks beside her at walking speed, 35mm lens, shallow depth of field, neon reflections on wet pavement.
• Multi-shot prompt (shot list)
  • Shot 1 (wide, 0-2s): Empty neon street at night, rain reflections.
  • Shot 2 (tracking, 2-5s): Woman walks into frame, camera follows.
  • Shot 3 (close-up, 5-7s): Slow push-in on her face, soft lighting.

Image Editing (Modify What You Already Have)

• Describe which part you want edited
• Describe what should change and what must stay the same
• Add details like lighting, mood, realism level
• Example: “Keep the room layout and style the same. Replace the blue chair with a modern white armchair. Maintain lighting, shadows, and floor reflections.”

Image Reference (Match style, lighting, camera, or objects)

Look, lighting, camera angle, color palette, texture, material, object shape.

• Upload your reference image
• Tell Seedream what to copy from it
• Tell Seedream what to generate differently
• Example: “Match the reference image’s camera angle and layout. Replace the object with a premium cola bottle, keeping the environment identical.”

Multi-Image Combination (Blend Images Together)

• Face from Image A + Body from Image B
• Environment from Image A + Object from Image B
• Style from Image A + Layout from Image B
• Upload 2–4 images and tell Seedream exactly what to take from each one
• Example: “Use Image 1’s environment and lighting. Use Image 2’s product design. Combine them into a single photoreal image with correct shadows.”

Multi-Image Output (Generate a Set of Matching Images)

• Match the same style, product, quality, and theme
• Useful for product catalogs, ad variations, character sheets, moodboards, and scene packs
• Tell Seedream: “Generate multiple variations of…” then describe what must stay consistent and what may change
• Example: Generate 6 matching images of the same young girl character. Keep her face and hair the same. Change outfits and environments

Tips

• Think like a commercial photographer
• Always define lighting direction
• Add texture details for realism
• Include subtle motion cues
• Preserve identity in Image-to-Image
• Change environment, not facial structure
• Keep composition intentional (clear focal point)
• Use cohesive color grading
• Control proportions and realism
• End with a quality enhancement block

Short Formula Version

Subject + Setting + Lighting + Mood + Action + Detail + Composition + Quality Flags

Example: Athletic founder standing in a modern training studio, soft side lighting, subtle dust in air, confident grounded stance, realistic fabric folds, shallow depth of field, ultra high resolution, high quality with realistic details, commercial photography realism, 4K.

Here’s what you can do:

• 1. Text-to-Image
  • Turn any description into a realistic or stylized image.
  • Example: “A nano banana dessert on a marble plate, warm candlelight, luxury vibe.”
• 2. Image + Text Editing
  • Upload an image and change or add anything. You can replace backgrounds, objects, or style.
  • Example: “Take this banana photo and replace the background with a tropical sunset.”
• 3. Multi-Image Blending & Style Transfer
  • Combine several images into a new scene or apply one image’s style to another.
  • Example: “Blend this banana photo with a watercolor painting style.”
• 4. Step-by-Step Refinement
  • Work gradually. Make small changes until the image is perfect.
  • Example: “Make the shadows softer, then add subtle highlights.”
• 5. Text in Images
  • Generate clear, readable text in images — perfect for logos, posters, or product labels.
  • Example: “Logo: ‘Nano Banana’ in bold gold letters on a dark background, banana icon above.”
• 6. High-Quality, Realistic Output — Supports 2K or 4K resolution, studio-style lighting, and realistic textures.
• 7. Consistency Across Scenes — Keep the same person, object, or style across multiple images or edits.
• 8. Smart Object & Scene Edits — Replace objects, change lighting, or adjust camera angle without losing details.

How to Structure Your Prompt

• Style & Art Direction — Photorealistic, hyper-detailed, Pixar-style, anime, vaporwave, cyberpunk, watercolor, digital illustration
• Scene Description — Describe the environment, atmosphere, and mood
• Main Subject (Hero Element) — Who or what is the focus? Person, object, or animal
• Camera & Cinematic Controls — Lens type, angle, perspective, composition, focal length
• Lighting — Soft, backlit, golden hour, volumetric rays, neon glow
• Texture, Color & Material — Matte, glossy, metallic, pastel, or vibrant textures
• Optional: Negative Prompts — What to avoid: blur, artifacts, extra limbs, wrong anatomy

Tips for Writing Better Prompts

• Be specific: “a cat” → weak; “a fluffy orange British Shorthair kitten sitting on a marble countertop, soft morning light” → strong
• Use layered descriptions: combine mood, materials, and atmosphere
• Use cinematic language: AI responds well to film terms
• Define lighting: one of the most important factors
• Avoid overloading the prompt: too many instructions can confuse the model
• Include references if you want consistent objects or people across images

Core Prompt Structure

A strong prompt usually includes:

• Task → What you want (write, summarize, generate an image, translate, etc.)
• Subject → Who/what it's about
• Style / Tone → Professional, playful, cinematic, etc.
• Format → Bullet points, blog post, table, list, etc.
• Details → Mood, setting, perspective, length, constraints

Example Prompts

Text Example

• ❌ "Summarize this text."
• ✅ "Summarize the following text in 3 bullet points, under 150 words, tone: professional."

Image Example

• ❌ "A wolf in a forest."
• ✅ "Photorealistic grey wolf standing in a misty forest at dawn, golden light, shallow depth of field, 16:9 aspect ratio."

Quick Tips at a Glance

• Always define task + style + format
• Use specific adjectives for images
• Refine instead of overloading prompt
• Generate multiple variations to compare
• Avoid contradictions (e.g. "minimalist but also highly detailed")

Why GPT Image 2 is a Major Upgrade

• “Thinking” Before Generating — Unlike older models, GPT Image 2 performs a reasoning step before creating images, improving composition, layout accuracy, and prompt understanding.
• Near-Perfect Text Rendering — Accurate text inside images (logos, UI, ads) across multiple languages. Ideal for ads, thumbnails, product labels, and UI mockups.
• High-Fidelity & Photorealism — Realistic lighting, textures, and materials with strong product rendering. Great for e-commerce visuals, marketing creatives, and cinematic thumbnails.
• Multi-Image + Reference Control — Supports multiple reference images (up to ~16) and maintains character / product consistency.
• Flexible Resolution (Up to 4K) — 1K / 2K / 4K output, any aspect ratio (portrait, landscape, square).
• Strong Instruction Following — Follows prompts more precisely and handles complex scenes easily.

Best Use Cases

• Product ads & e-commerce visuals
• Social media creatives
• Thumbnails with text
• UI / app mockups
• Posters & marketing designs
• Storyboards & cinematic scenes

Prompt Structure

[Main Subject] + [Environment / Scene] + [Style / Look] + [Lighting] + [Camera / Composition] + [Text (if needed)]

Key Prompting Rules

• Be specific (no vague prompts)
• Describe text in quotes → "like this"
• Mention lighting & camera
• Keep 1 clear idea (no chaos)
• Avoid too many styles in one prompt, random keywords, or overloading details

How Flux 2 Works

• Describe Only What You Want — Flux 2 follows instructions very literally. Avoid extra descriptions of things you don't want.
  Example: "Crisp product shot on a matte surface, clean lighting, smooth background."
• Use Compact, Direct Style Tags — Flux 2 prefers short, strong stylistic labels. Use one or two style tags only — too many creates conflict.
• Anchor the Scene With Specific Composition — Flux 2 responds extremely well to camera and framing cues.
  • "Center-framed hero object"
  • "Low-angle shot"
  • "Symmetrical composition"
• Material & Surface Precision — Flux 2 is highly sensitive to textures. Specify surface material, color palette, and lighting behavior.
  Example: "Reflective metallic surface with soft specular highlights, charcoal-black palette."
• Control Lighting Intentionally — Lighting can dramatically change the output.
  • "Soft diffused lighting"
  • "Studio three-point lighting"
  • "Golden-hour backlight"
• HEX Color Control — Flux 2 fully supports precise hex color values for exact branding and product accuracy.
  • "Product label highlight in #FF5733"
  • "Warm tone accents in #D4A373"
• Typography & Design — FLUX.2 generates clean typography, product marketing materials, and magazine layouts.
  Example: "product advertisement, ‘Ultra-strong titanium’ headline, ‘Shielded in a strong titanium frame, your Galaxy S25 Ultra always stays protected’ subtext, close-up of phone edge showing titanium frame, dark gradient background, clean minimalist tech aesthetic, professional product photography"
• Multilanguage Prompting — Flux 2 supports multi-language prompting. You can write prompts in Hindi, English, Korean, Japanese, Arabic, Spanish, and more.
  • Korean: "따뜻한 조명 아래 고급 스킨케어 병의 스튜디오 샷."
  • Spanish: "Retrato cinematográfico con luz suave y detalles realistas."
• Character & Object Control — For Image-to-Image or Image Elements prompts, be specific:
  • "Keep face identical"
  • "Preserve hairstyle"
  • "Match object proportions"

Beside AI video tools, you can also generate native sound effects and lip sync for the videos to have more immersive experience. You can simply add “Audio:” followed by the desired sound to the prompt itself (Make sure to have Audio turned on in the Apps or via the API).

Elevenlabs V3 TTS