Awesome Sora 2 Prompts
repository·main·Indexed 42 days ago
https://github.com/zhangchenchen/awesome_sora2_promptA curated resource for prompt engineering tailored for OpenAI's Sora and Sora 2 video generation models. It features structured frameworks like the Five Pillars, guides on cinematic techniques (shot scales, camera angles, and movements), and collections of official and viral community prompts. The documentation provides strategies for using Sora as a 'world simulator', including guidance on synchronized audio, enhanced physics, and prompt length optimization for creative exploration versus precision execution.
What's inside Awesome Sora 2 Prompts
- This repository is a curated collection of high-quality prompts designed for OpenAI's Sora and Sora 2 video generation models. It provides deep analysis and practical guidance for prompt engineering to help users leverage Sora's capabilities as a 'world simulator'.
What is Sora and Sora 2?
mainSora is OpenAI's text-to-video generation model. While the original Sora focuses on realistic and imaginative video creation, Sora 2 (released September 30, 2025) introduces several advanced capabilities:
- Synchronized Audio: Generation of natural sound effects and dialogue.
- Enhanced Physics: Improved realism in motion and object interactions.
- Self-Insertion Cameos: The ability to insert custom avatars into videos.
- Advanced Controllability: More precise camera work and cinematic control.
Best practices for Sora 2 prompt engineering
mainBased on community analysis of over 32 experiments, follow these insights to optimize video generation results:
Recommended
- Keep it concise: Prompts under 120 words tend to perform better for creative content.
- Use camera instructions: Specific camera angles and movements dramatically improve visual impact.
- Focus on animals: Animal-focused prompts show high success rates.
- Add sensory details: Include specific textures, lighting, and reflections to enhance realism.
Avoid
- Overly complex physics: Extremely detailed physics instructions may cause visual artifacts.
- Vague descriptions: Ambiguous language leads to inconsistent video results.
How to use the 'World Simulator' Paradigm in prompting
mainSora operates as a 'world simulator' rather than a simple video generator. When prompting, you should treat the scene as a self-consistent miniature world by:
- Providing initial conditions and physical laws.
- Specifying how objects interact with each other.
- Defining environmental properties such as gravity, lighting, and reflections.
- Guiding the physics either implicitly or explicitly.
Understand the World Simulator Mindset for Sora
mainWhen prompting Sora, avoid simply listing visual elements (e.g., "coffee cup, steam, wood"). Instead, treat Sora as a world simulator by describing a world with internal logic, physical properties, and environmental context.
Key Principles:
- Physics: Describe how objects interact (e.g., "steam rises from the hot liquid").
- Environment: Provide context (e.g., "morning sunlight streaming through a window").
- Dynamics: Include movement and material interactions (e.g., "liquid ripples gently").
✅ DO: Describe a world with internal logic "A coffee cup sits on a wooden table. Steam rises from the hot liquid, creating swirling patterns that catch the morning sunlight streaming through a nearby window. The liquid ripples gently." ❌ DON'T: Just list visual elements "coffee cup, steam, wood, window, sun"Universal principles for effective Sora prompting
mainWhen constructing prompts for Sora, follow these core principles to ensure high-quality output:
- Core Components: Always include a clear subject, action, and environment.
- Specificity: Use specific descriptors rather than vague adjectives.
- Physics Guidance: Guide physical interactions by describing materials and their behaviors.
- Complexity Matching: Match the complexity of your prompt to your goal—use simpler prompts for creative exploration and highly detailed prompts for precision execution.
Emerging Prompt Patterns in Sora 2
mainSora 2 prompts have evolved from the highly detailed descriptions of Sora 1 toward more diverse and layered structures. Key emerging patterns include:
- Audio-First Prompts: Explicitly requesting sounds, dialogue, or music.
- Style Shorthand: Using single phrases (e.g., "South Park", "luxury ad") to trigger complex styles.
- Sequential Specifications: Breaking down transformations or actions step-by-step.
- Nested Realities: Creating "screens within screens" or simulations within simulations.
- Real-World References: Using celebrity names, brands, or current events to ground the content.
How Sora 2 prompting differs from Sora 1
mainSora 2 prompting shifts from scene description to complete technical specification. While Sora 1 prompts focus on describing the subject, action, and environment, Sora 2 prompts utilize a structured, multi-dimensional approach that mimics a professional film production pipeline.
Key innovations in Sora 2 prompting include:
- System-Level Specification: Controlling how a scene is captured, not just what is in it.
- Technical Precision: Using industry-standard terminology for camera gear (e.g.,
gimbal-stabilized,aspherical 50mm lens), filters (e.g.,1/8 Black Pro-Mist filter), and post-production (e.g.,color grading,grading & palette). - Audio Integration: Explicitly requesting synchronized dialogue or environmental sounds (e.g.,
calling out loudly). - Structured Sections: Organizing prompts into logical blocks like
Primary Target & Visuals,Format & Appearance,Camera & Filter,Grading & Palette, andLighting & Atmosphere.
Apply Lighting Techniques to Sora Prompts
mainUse specific lighting descriptions to define the mood, atmosphere, and visual quality of a scene. Common techniques include:
- Golden Hour: Warm, soft light (sunrise/sunset) for romantic or nostalgic moods.
- Blue Hour: Cool twilight for mysterious or ethereal moods.
- High-Key Lighting: Bright, even lighting with minimal shadows for optimistic or professional aesthetics.
- Low-Key Lighting: High contrast and deep shadows for dramatic or film noir moods.
- Rembrandt Lighting: A specific 45-degree angle setup creating a triangle of light on the cheek for artistic portraits.
- Practical Lights: Using visible light sources (lamps, neons) for a realistic, lived-in atmosphere.
- Volumetric Lighting / God Rays: Visible light beams through atmosphere for a divine or spiritual effect.
- Silhouette: Backlighting a subject to create a dark, mysterious shape against a bright background.
Golden hour cinematography of a couple walking on the beach. Low sun creates long shadows, warm orange and pink tones wash over everything. Soft, flattering light wraps around subjects, creating magical, romantic atmosphere.Best Practices for Hyperrealistic Landscape Prompts
mainTo achieve maximum visual fidelity in landscape generation, follow these guidelines:
Do's ✅
- Specify Lighting Conditions: Use terms like
golden hour,blue hour,overcast, orharsh midday. - Include Atmospheric Elements: Add
fog,mist,dust,rain, orsnow. - Define Camera Movement Precisely: Specify
speed,direction, andsmoothness(e.g.,steadicam,drone view,FPV). - Add Dynamic Elements: Include movement like
wind,water,clouds, orwildlife. - Specify Film Stock or Digital Format: Use technical terms like
35mm,IMAX, orcinema camera. - Use Reference Locations: Mentioning specific places helps the model understand the intended aesthetic.
- Include Reflections and Shadows: Explicitly mention these to guide light physics.
Don'ts ❌
- Don't Forget Scale: Always include scale references to prevent landscapes from feeling flat.
- Don't Ignore Time of Day: Lighting is critical for mood.
- Don't Use Static Descriptions: Ensure there is movement (wind, water, or camera movement).
- Don't Overload with Detail: Focus on 3-5 key elements maximum to maintain prompt coherence.
- Don't Forget Depth: Explicitly define
foreground,midground, andbackgroundto create dimension.
- Specify Lighting Conditions: Use terms like
Explore Official OpenAI Showcase Prompts
mainThe repository provides a collection of 'gold standard' prompts used by OpenAI to demonstrate Sora's capabilities. These include high-fidelity examples like 'Tokyo Walk', 'Wooly Mammoths', and 'Dragon over Glacier' (Sora 2). For each showcase prompt, you can view the video and access a detailed analysis of the prompt structure in the dedicated documentation file.
<!-- To see detailed analysis for a specific prompt, navigate to: --> <a href="./prompts/official-prompts.md#tokyo-walk">📖 Details</a>Apply Advanced Camera Movements and Effects
mainEnhance the dynamic nature of a scene using advanced cinematic techniques:
- Dolly Zoom (Vertigo Effect): Physically moving the camera back while zooming in (or vice versa) to distort background perspective and create unease.
- Speed Ramping: Smoothly transitioning between normal speed and slow/fast motion to emphasize action.
- Whip Pan: An extremely fast pan that creates motion blur, used for energetic transitions.
Dolly zoom effect on character having sudden realization. Camera physically pulls back while simultaneously zooming in, keeping subject same size but distorting background perspective. Creates unsettling, dramatic Vertigo effect.