Building a Prompt Enhancement Library for AI Image and Video Generation
If you’ve spent any time with AI image or video tools, you’ve probably noticed something:the difference between an average result and a stunning one usually comes down to the prompt more like visualgpt
Anyone can type “a cat in a city.”
But creators who understand how prompts work under the hood get results that feel intentional, cinematic, and polished.
That’s where a Prompt Enhancement Library comes in.
Yes, absolutely—we can create (and maintain) a shared library or glossary of technical descriptors specifically for AI prompts. This is a proven concept in prompt engineering: by incorporating precise, industry-standard terms from photography, cinematography, and digital art, users can guide models like Sora AI (for videos) or tools like DALL-E/Midjourney (for images) to produce higher-quality, more consistent outputs. These aren’t just random words; they’re borrowed from real-world creative workflows to add depth, realism, and control.
The idea is simple: Anyone can start with a basic prompt like “a cat in a city,” but adding descriptors elevates it to “a fluffy tabby cat prowling a neon-lit cyberpunk alley at dusk, low-angle wide shot, shallow depth of field, dramatic volumetric lighting with blue-purple palette, 8K resolution, cinematic style inspired by Blade Runner.” This results in sharper details, better composition, and more engaging visuals.
Based on best practices from OpenAI’s Sora 2 guides, advanced AI photo techniques, and creative glossaries, here’s a starter Prompt Enhancement Library. I’ve categorized it for ease—pick terms from relevant sections and plug them into your prompts. This is tailored for Sora (via Perplexity/Comet Browser) but works broadly for text-to-image/video tools. Start small: Add 2-3 descriptors per category to avoid overwhelming the AI.
Pro Tip Library (Copy-paste these into your Comet prompts):
- Lighting: “volumetric god rays”, “golden hour soft light”, “dramatic rim lighting”, “teal-orange color grade”
- Camera: “slow dolly-in tracking shot”, “low-angle wide establishing shot”, “shallow depth of field with creamy bokeh”
- Style: “hyper-realistic 8K cinematic”, “35mm film grain”, “Blade Runner cyberpunk aesthetic”
- Motion (videos): “smooth parabolic arc”, “slow-motion splash”, “diegetic ambient sounds with orchestral swell”
Example Prompt to Try Now in Sora: “Generate a Sora video: A sleek cyberpunk drone flying over a rainy neon city at night, low-angle tracking shot, volumetric lighting with god rays piercing fog, teal-orange palette, hyper-realistic 8K, dramatic synth score.”
Then ask Comet Assistant: “Post this video to my X with caption ‘My first Sora masterpiece! 🌃 #AISora #CometBrowser’”
Download Comet (free) and upgrade to Max for unlimited Sora 2 Pro—it’s the easiest way to create and share viral AI videos right now. Who’s trying this today? 👇
1. Camera Movements & Framing (Adds dynamic perspective; great for videos to control flow)
- Wide establishing shot: Sets a broad scene overview.
- Medium close-up: Focuses on subject details without full body.
- Low-angle / High-angle: Makes subjects appear powerful (low) or vulnerable (high).
- Tracking left/right: Camera follows movement horizontally.
- Slow dolly-in / Push-in: Gradually zooms toward the subject for emphasis.
- Pan / Tilt: Sweeps side-to-side or up-down.
- Aerial wide shot: Bird’s-eye view for epic scale.
- Handheld: Adds shaky, realistic documentary feel.
- Over-the-shoulder: Views action from behind a character.
- Crane shot: Smooth elevation for dramatic reveals.
Example Boost: Basic: “A runner in a park.” Enhanced: “A runner sprinting through a misty park at dawn, low-angle tracking shot following from behind.”
2. Lighting & Palette (Improves mood, realism, and color depth; key for hyper-realistic outputs)
- Volumetric lighting: Creates beams/god rays through fog or dust.
- Golden hour / Blue hour: Warm sunset glow or cool pre-dawn tones.
- Dramatic side lighting: Strong shadows from one side for contrast.
- Soft rim light / Backlight: Outlines subjects with a glowing edge.
- Warm tungsten / Cool LED: Indoor lamp hues vs. modern blueish lights.
- HDR (High Dynamic Range): Balances bright highlights and deep shadows.
- Teal-orange palette: Cinematic color grade (cool shadows, warm skin).
- Negative fill: Darkens one side for moodier depth.
- Bokeh effect: Blurry background lights as circular orbs.
- Ambient bounce: Soft reflected light from surfaces.
Example Boost: Basic: “A forest scene.” Enhanced: “Ancient forest at twilight, volumetric lighting filtering through leaves, teal-orange palette with soft rim light on trees.”
3. Styles & Aesthetics (Sets overall vibe; references real artists/films for consistency)
- Photorealistic / Hyper-realistic: Lifelike detail, like a photo.
- Cinematic: Movie-like with flares, grain, and epic scale.
- Cyberpunk: Neon, rainy, futuristic dystopia.
- Vintage 35mm film: Grainy, warm, with light leaks.
- Renaissance oil painting: Rich textures, dramatic chiaroscuro.
- Documentary-style: Raw, unpolished, like handheld footage.
- Ethereal / Surreal: Dreamy, floating elements.
- Blade Runner aesthetic: Moody sci-fi with reflections.
- Annie Leibovitz-inspired: Dramatic portraits with bold lighting.
- Stop-motion feel: Tactile, jerky animation.
Example Boost: Basic: “A spaceship landing.” Enhanced: “Futuristic spaceship descending on an alien planet, cyberpunk aesthetic with neon flares, cinematic 35mm film style.”
4. Composition & Depth (Enhances visual structure; prevents flat or cluttered results)
- Rule of thirds: Places key elements off-center for balance.
- Shallow depth of field (DOF): Sharp subject, blurred background.
- Symmetrical composition: Mirrored elements for harmony.
- Leading lines: Paths or edges guiding the eye.
- Anamorphic lens: Wide, stretched look with oval bokeh.
- Maximum depth of field: Everything in sharp focus.
- Parallax effect: Layered foreground/background movement.
- Vignette: Darkened edges to focus center.
- Foreground elements: Adds layers (e.g., branches framing the scene).
- Bokeh background: Soft, out-of-focus blur.
Example Boost: Basic: “A portrait of a woman.” Enhanced: “Elegant woman in a garden, rule of thirds composition, shallow DOF with creamy bokeh background.”
5. Actions & Timing (For videos: Adds precise motion; keeps outputs coherent)
- Slow-motion: Emphasizes details in key moments.
- In beats: Breaks action into steps (e.g., “takes three steps, pauses, jumps”).
- Flares / Twirls: Dynamic subject movements.
- Transitions: “Fade to” or “cut to” for multi-shot.
- Timestamps: [00:00-00:05] for sequenced clips.
- Diegetic movement: Natural, in-world actions like footsteps.
- Parabolic arc: Smooth curving paths (e.g., for throws).
- Flinch / Widen eyes: Subtle emotional reactions.
- Coiling / Splashing: Fluid, physics-based effects.
- Hiss / Crackle: Onomatopoeic for sound-synced actions.
Example Boost: Basic: “A chef cooking.” Enhanced: “Chef drizzling honey over yogurt in slow-motion, coiling beautifully, [00:00-00:04] close-up on spoon.”
6. Sound & Atmosphere (Video-specific: Enhances immersion; Sora 2 supports synced audio)
- Diegetic sounds: In-scene (e.g., rain patter, footsteps).
- Non-diegetic score: Background music (e.g., orchestral swell).
- Ambient hum / Murmur: Subtle background noise.
- Percussive beat: Rhythmic audio for action.
- Dialogue block: Labeled speakers (e.g., “Character: ‘Hello world.’”).
- Muffled announcement: Distant, echoed effects.
- Wind whistle / Fabric flutter: Environmental audio.
- No score: Pure natural sounds for realism.
- LUFS level: Audio volume spec (e.g., -20 LUFS for balanced).
- Hiss of machine / Clinking cups: Specific FX.
Example Boost: Basic: “A cafe scene.” Enhanced: “Busy cafe at morning rush, ambient hum of espresso machines and murmurs, diegetic only—no score.”
7. Technical Specs (Boosts quality/resolution; mimics pro equipment)
- 8K / 16K resolution: Ultra-sharp detail.
- 50mm / 85mm lens: Standard portrait focal lengths.
- f/1.4 aperture: Wide for low-light bokeh.
- Canon EOS R5 / Hasselblad: Camera references for realism.
- Film grain / Halation: Vintage glow on highlights.
- Black Pro Mist filter: Soft diffusion.
- Volumetric fog: 3D atmospheric effects.
- HDR imaging: Wide color range.
- Spherical primes: Clean, non-distorted lenses.
- CPL rotation: Polarizer for reduced glare.
Example Boost: Basic: “A mountain view.” Enhanced: “Snow-capped mountains at dawn, 8K resolution, shot on 50mm f/1.8 lens with film grain and HDR.”
How to Use This Library Effectively
- In Comet Browser/Perplexity: For Sora, start a prompt with “Generate a Sora video of…” and layer in 4-6 descriptors. Use the AI sidebar to remix: “Remix this with added shallow DOF and golden hour lighting.”
- Tips: Be specific but concise—overloading can confuse the AI. Test iterations: Generate, then tweak one category at a time. For images, focus on static elements (lighting/composition); for videos, add motion/sound.
- Expanding It: We could crowdsource this via a shared doc or Comet community—users add tested terms (e.g., from Midjourney discords). Reference guides like OpenAI’s Sora Cookbook for updates.
- Why It Works: These terms leverage the model’s training on real media, tricking it into higher fidelity. Studies show detailed prompts increase output scores by 20-50% in realism and coherence.
This library is a starting point—try it out and refine based on results! If you share a base prompt, I can enhance it with these.