Foreword
As the new-generation AI video generation model from ByteDance, Seedance 2.0 is redefining the boundaries of AI video creation with its powerful multimodal understanding, exceptional character consistency, and precise execution of complex camera language. It is no longer a simple "text-to-video" tool, but a powerful platform that grants creators "director-level" control. However, to truly master this tool and unleash its full potential, the key lies in mastering its unique prompt engineering methodology.
This guide aims to provide you with a comprehensive, in-depth, and highly practical bible for Seedance 2.0 prompts. We will systematically analyze its working principles, deconstruct an effective prompt formula, and provide a massive case library with over a hundred examples, covering various styles from movies, advertisements, and anime to social media. Whether you are a beginner or a professional seeking advancement, this guide will help you transition from "random generation" to "precise direction," creating stunning, cinematic-quality AI videos.
Chapter 1: Core Concepts - Understanding the "Director's Mindset" of Seedance 2.0
Before writing your first prompt, you must understand the fundamental difference between Seedance 2.0 and other models: it interprets instructions with a "director's mindset," not just by recognizing keywords. This means your prompt needs to be like a detailed shooting script, clearly conveying your creative intent.
1.1 Multimodal Fusion: A Creative Dimension Beyond Text
The core advantage of Seedance 2.0 lies in its powerful multimodal input capabilities. It allows you to use text, images, video, and audio simultaneously as creative materials, greatly enhancing control and precision.
- Text: The core instruction, used to describe all elements such as scenes, actions, and styles.
- Image: Can be used to lock in character appearance, scene style, composition, and even define the first and last frames of the video, allowing the AI to automatically generate the transition.
- Video: Primarily used as a reference for camera movement, character dynamics, and editing rhythm. The AI learns the cinematography and action logic from the reference video.
- Audio: Used to drive the rhythm and mood of the video, and can even achieve precise lip-sync.
In your prompt, you can use the @ symbol to reference these uploaded materials, such as @image1 for the first reference image and @video1 for the reference video. This "learning by example" model is the key to achieving advanced control.
1.2 Key Capabilities: Why is Seedance 2.0 So Powerful?
-
Character and Scene Consistency: This is one of Seedance 2.0's most groundbreaking features. By using reference images or creating character profiles, the model can maintain a high degree of consistency in a character's face, clothing, and hairstyle across multiple shots and scenes, solving the common problems of "face-swapping" and "flickering" in AI video creation.
-
Physics Realism: The model has a deep understanding of the physical world, enabling it to generate videos that conform to the logic of gravity, fluid dynamics, and lighting. Whether it's the flutter of cloth, the splash of liquid, or the force feedback in a fight, the realism is remarkable.
-
Multi-Shot Storytelling: You can orchestrate a short film with multiple shots in a single prompt using timestamps (e.g.
[00-05s]) or scene descriptions. The AI automatically handles transitions and connections between shots, achieving a preliminary form of "directorial thinking."
Chapter 2: Anatomy of a Perfect Prompt - The Core Formula and Structure
Through the collective experience of many practitioners, an effective structure for Seedance 2.0 prompts has emerged. It breaks down complex creative needs into modular components, ensuring the AI can understand and execute them accurately and orderly.
2.1 The Ultimate 8-Part Formula
This is a comprehensive, "universal formula" that covers all the elements required for most high-quality videos:
Subject + Action + Scene + Lighting + Camera Language + Style + Quality + Constraints
| Component | Description | Key Points & Examples |
|---|---|---|
| Subject | The core character or object in the frame | Be clear and specific. E.g. "A young man in a black trench coat" is better than "a person." |
| Action | What the subject is doing | Use continuous, natural, slow-paced verbs. E.g. "Slowly turns around, gently raises a hand" is better than "dances." |
| Scene | The environment where the subject is located | Describe the specific location, time, and atmosphere. E.g. "A rainy Tokyo street at night, neon lights flickering." |
| Lighting | The lighting effect of the scene | Match the style. E.g. "Soft warm light of dusk," "Cold blue and purple tones of cyberpunk." |
| Camera | The camera's perspective, movement, and focal length | This is the key to achieving a "cinematic feel." E.g. "Slow push-in close-up shot," "Handheld camera following shot." |
| Style | The overall visual and artistic style of the video | Can be a director, art movement, or specific period. E.g. "Wong Kar-wai film style," "Van Gogh post-impressionist oil painting style." |
| Quality | The resolution and texture of the video | Use high-quality terms directly. E.g. "4K, ultra HD, cinematic quality, rich in detail, sharp." |
| Constraints | Explicitly what not to do; negative prompts | Used to avoid common errors. E.g. "Stable face without distortion, no stiff movements, no watermarks." |
2.2 The Streamlined 4-Part Structure
For quick creations or relatively simple scenes, you can use:
Subject → Action → Camera → Style
This structure emphasizes the logical sequence of storytelling: Who (Subject) is doing what (Action), how we see it (Camera), and what feeling it presents (Style).
2.3 Comprehensive Master Prompt Template
Combining the advantages of the two structures above, here is a master template you can copy and use directly:
[Subject]: [Describe the subject clearly, including age, clothing, material, etc.]
[Action]: [Use single or continuous, present-tense verbs to describe the core action of the subject]
[Scene]: [Describe the time, place, environment, atmosphere, and key background elements]
[Lighting]: [Describe the key light, ambient light, color tone, e.g. Rembrandt lighting, neon lights, cool tones]
[Camera]: [Shot Size: e.g. close-up, medium shot, full shot] + [Movement: e.g. push, pull, pan, track, follow] + [Angle: e.g. low-angle, high-angle] + [Lens: e.g. wide-angle, telephoto]
[Style]: [A core visual anchor: e.g. film, art movement, director's name] + [Color Grading: e.g. high saturation, faded, black and white]
[Quality]: [4K, Ultra HD, cinematic quality, rich in detail, no noise]
[Constraints/Negative Prompts]: [Consistent face, consistent clothing, smooth and non-stiff movements, no flickering, no watermarks, no blur]Chapter 3: The Prompt Vocabulary - Building Your "Director's Toolkit"
3.1 Camera Language Vocabulary
| Category | Term (English) | Description & Effect |
|---|---|---|
| Shot Size | Close-up | Focuses on facial or object details, emphasizing emotion. |
| Medium Shot | Shows the upper body and some of the environment, balancing character and background. | |
| Full Shot / Long Shot | Shows the entire body and its environment, establishing relationships. | |
| Wide Shot / Establishing Shot | Establishes the scene, shows a grand background, characters are small. | |
| Camera Angle | Low-angle Shot | Shoots from bottom to top, making the subject appear tall and powerful. |
| High-angle Shot | Shoots from top to bottom, making the subject appear small and weak. | |
| Bird's-eye View | A vertical overhead shot, often used to show geographical patterns or abstract compositions. | |
| Dutch Angle / Canted Angle | A tilted shot, creating a sense of instability, tension, or madness. | |
| Camera Movement | Dolly in / Push in | The camera moves closer to the subject, gradually focusing, emphasizing importance. |
| Dolly out / Pull out | The camera moves away from the subject, revealing a wider environment, often used for endings. | |
| Pan | The camera rotates horizontally, used to follow a moving subject or show a wide scene. | |
| Track / Truck | The camera moves parallel to the subject, often used to follow walking or moving vehicles. | |
| Crane / Boom | The camera moves vertically up or down, changing the viewing height. | |
| Orbit / 360-degree shot | The camera moves in a circle around the subject, showing it from all angles. | |
| Handheld | Simulates the slight shaking of a handheld camera, increasing realism and immersion. | |
| Steadicam / Gimbal | The footage is extremely smooth and fluid, suitable for elegant, professional motion shots. |
3.2 Lighting & Color Vocabulary
| Category | Term (English) | Description & Effect |
|---|---|---|
| Light Quality | Soft Light | Diffused light, soft shadows, suitable for beautiful, healing styles. |
| Hard Light | Direct light, sharp shadows, suitable for emphasizing drama and tension. | |
| Light Direction | Rembrandt Lighting | Classic portrait lighting with a triangular patch of light on one side of the face. |
| Butterfly Lighting | The main light is directly above the subject, creating a butterfly-shaped shadow under the nose. | |
| Backlight / Rim Light | Light shines from behind the subject, creating a contour light that separates the subject from the background. | |
| Color Style | Cinematic Color Grading | Usually refers to the Teal and Orange look, with strong warm-cool contrast. |
| High Saturation | Bright, vibrant colors, suitable for advertisements and animations. | |
| Desaturated / Faded | Muted colors, creating a retro, melancholic, or sophisticated feel. | |
| Monochrome / Black and White | Strips away color to emphasize light, shadow, composition, and emotion. | |
| Warm Tones | Dominated by red, orange, and yellow, feels warm, nostalgic, and comfortable. | |
| Cool Tones | Dominated by blue, green, and purple, feels calm, distant, and technological. |
3.3 Style & Genre Vocabulary
| Category | Term / Director | Key Elements |
|---|---|---|
| Film Directors | Wong Kar-wai | Neon lights, rainy nights, slow motion, emotional colors, retro feel. |
| Wes Anderson | Symmetrical composition, high-saturation warm tones, retro props, flat aesthetic. | |
| Denis Villeneuve | Epic scale, huge contrasts in scale, minimalist composition, oppressive color tones. | |
| Quentin Tarantino | Non-linear narrative, aestheticized violence, retro soundtracks, strong dialogue scenes. | |
| Art Movements | Cyberpunk | Neon lights, towering skyscrapers, rainy nights, futuristic technology, Eastern elements. |
| Van Gogh Style | Thick impasto brushstrokes, swirling starry skies, strong blue-yellow contrast. | |
| Ink Wash Painting | Primarily black and white, negative space, profound mood, Eastern aesthetics. | |
| Video Types | Movie Trailer | Fast-paced editing, grand scenes, suspense, text titles. |
| Commercial Ad | Product close-ups, bright lighting, high saturation, positive atmosphere. | |
| Documentary | Realistic feel, handheld shots, natural light, interview shots. | |
| UGC Style | Vertical phone shooting, everyday scenes, imperfect realism. |
3.4 Negative Prompts & Constraints Vocabulary
| Category | Common Constraints / Negative Prompts |
|---|---|
| Character Consistency | consistent face, no distortion, clear facial features, consistent clothing, consistent hairstyle |
| Motion Fluency | smooth and natural motion, not stiff, physically accurate, no sudden jumps |
| Image Quality | no blur, no noise, no flickering, no watermark, no text overlay, stable footage |
| Composition & Aesthetics | avoid weird composition, avoid clipping, correct proportions, no deformation |
| Content Safety | non-violent, not explicit, safe for work |
Pro Tip: Place the most critical constraint words at the end of your prompt as a "quality defense line." For example, almost all character videos should include: consistent face, no distortion, consistent clothing, smooth and natural motion.
Chapter 4: The Ultimate Case Library - Selected Prompt Case Studies
4.1 Cinematic & Narrative Styles
Case 1: A Rainy Night Phone Booth in Wong Kar-wai Style
- Creative Goal: Replicate the classic atmosphere of 1990s Hong Kong art-house films, emphasizing emotion and visual style.
- Key Techniques: Clearly specify the director's style, film grain, color tone, and the unique slow-shutter drag effect. Use "core dialogue" to anchor the character's emotional tone.
[Style]: 90s Hong Kong Art Cinema style, retro film feel, high ISO grain, ambiguous yellow-green tint, frame stepping effect, melancholic atmosphere.
[Core Dialogue (for emotional control)]: "If memories could be canned, I hope this can would never expire."
[Video Duration]: 10 seconds
[Script]:
[00:00-00:04]: Shot 1 - Peeking through the glass. Scene: A red public phone booth covered in rain. Character: A man in a khaki trench coat, gripping the receiver, not speaking, just listening. Emotional Performance: Through the refraction of the glass, his empty but affectionate eyes are visible. Rain streams down the glass, distorting his face like an oil painting.
[00:04-00:07]: Shot 2 - Extreme close-up and micro-expressions. Focus on the character's lips and half of his face. He whispers softly into the receiver, his lips trembling slightly. Lighting: The neon lights from the street flow across his face, alternating between light and shadow.
[00:07-00:10]: Shot 3 - The iconic slow-shutter drag. The character hangs up the phone and turns to walk into the rainy crowd. Using a frame stepping effect, the character's back becomes blurry with motion blur. Environment: The background is flowing city car lights, forming long light trails.
[Technical Parameters]: Simulated handheld camera, shallow depth of field, color shift, emotionally rich.Case 2: Epic Desert Chase in Denis Villeneuve Style
[Style]: IMAX 70mm Film, Denis Villeneuve Style, Gritty Realism, Epic Scale, Desaturated.
[Video Duration]: 15 seconds
[Script]:
[00-05s]: A massive sandstorm, miles high, engulfs a vast desert. A tiny convoy of armored vehicles is fleeing from it. The contrast in scale between nature and humanity is terrifying. A sense of a tense, Hans Zimmer-style score.
[05-10s]: Inside the lead vehicle. The driver screams "Go! Go!" The camera shakes violently. Sand batters the windshield. The sun is obscured by the approaching wall of dust.
[10-15s]: The vehicle hits a huge dune and flies into the air (slow motion). It becomes a silhouette against the dark storm. Lightning flashes within the dust cloud. Debris flies past the camera. Cut to black on impact.Case 3: Hollywood Professional-Grade Racing Movie
[Style]: Hollywood Professional Racing Movie (Le Mans Style), Cinematic Night, Rain, High Stakes Sport.
[Video Duration]: 15 seconds
[Script]:
[00-05s]: Shot 1 - The Veteran (In-car/Close-up). Rain lashes the windshield of a high-tech race car. The veteran driver, wearing a helmet, watches calmly and focused. The dashboard lights reflect on his visor. Dialogue Cue: He gives a slight nod, mouthing "Let's go."
[05-10s]: Shot 2 - The Challenger (In-car/Close-up). Cut to the rival car next to him. The young driver grips the steering wheel, breathing heavily. His eyes are wide with adrenaline. Dialogue Cue: He whispers to himself, "Focus."
[10-15s]: Shot 3 - The Green Light (Wide-angle action). The starting lights turn green. Both cars accelerate in perfect sync on the wet asphalt. A huge splash of water hits the camera. Motion blur turns the stadium lights into long streaks of color.4.2 Advertising & Commercial Branding
Case 4: MUJI Brand Promotional Video
[Prompt]: Help me generate a promotional video about the MUJI brand.Case 5: Perfume Motion Graphics Ad
[Prompt]: Generate ad content for the perfume product in @image2 based on the script in @image1. The voiceover should be a natural female English voice. Pay attention to the proportion of the perfume bottle and blend it into the background with natural lighting. Avoid heavy texture overlays and cutout effects. The rhythm can be more upbeat.
[Reference Materials]: @image1: An image containing the animation script or storyboard. @image2: A high-definition image of the perfume product.4.3 Social Media & Viral Memes
Case 6: Giant Orange Cat Meme
[Style]: Mockumentary, Phone Vlog perspective, hyper-realistic CG combined with live-action, 8K quality, perfect fur physics simulation.
[Video Duration]: 15 seconds
[Scene]: Chongqing's Hongyadong or a busy interchange (with a magical 8D city feel).
[Script]:
[00:00-00:05]: Shot 1 - The Reveal. The scene shows a busy city street. The camera pans up to reveal a Godzilla-sized orange cat stuck between two skyscrapers. The giant cat is stuck because it's too fat, waving its huge paws, looking pitiful.
[00:05-00:10]: Shot 2 - The Interaction. The giant cat lowers its head, curiously sniffing a bus waiting at a red light. The bus driver calmly reaches out and pets the giant cat's nose. The cat sneezes, blowing away the leaves on the roadside trees and the hats of pedestrians.
[00:10-00:15]: Shot 3 - The Punchline. The giant cat finally squeezes through the buildings and sits down on a cross-river bridge, causing the bridge deck to sink slightly. It lazily lies down and starts grooming itself, blocking the entire evening rush hour traffic.4.4 Anime & Animation Styles
Case 8: Van Gogh Post-Impressionist Style Animation
[Style]: Van Gogh Post-Impressionism oil painting, Heavy Impasto texture, signature swirling brushstrokes, dreamy feel, high saturation blue-yellow contrast.
[Video Duration]: 15-second animation
[Visual Content]: This is a dynamic world made entirely of thick oil paint. Sky: In the deep blue night sky, huge yellow celestial bodies and a crescent moon are surrounded by short, radial brushstrokes. The nebula swirls wildly in the sky like a rushing river. Foreground: On the left is a huge cypress tree, twisting like a black flame, reaching for the sky. Background: The sleeping town in the valley, with warm, roundly painted yellow light coming from the windows of the houses. The entire scene slowly flows and breathes with the direction of the brushstrokes.4.5 Appendix: Extended Casebook Quick Reference
Case 11: Cyberpunk City Stroll
A female cyber-warrior in a glowing exoskeleton walks slowly through the rainy streets of Neo-Tokyo at night.
Camera: Low-angle shot, following shot, camera stays about 2 meters behind her.
Scene: Towering skyscrapers covered in holographic ads, neon lights reflecting on the wet streets, bustling crowds.
Style: Cyberpunk aesthetic, teal and orange tones, misty rain, cinematic quality.
Quality: 4K, ultra HD, rich in detail.
Constraints: Clear face, smooth motion, no flickering.Case 12: Ancient Xianxia Sword Flight
A female cultivator in white robes stands on a misty mountain peak, holding a long sword, and flies on her sword.
Camera: Zooms in from a distance, finally stopping at a medium shot of her front, her clothes and hair flying in the wind.
Scene: Fairyland-like mountains, rolling sea of clouds, faint palaces in the distance.
Style: Chinese Xianxia aesthetic, ink wash rendering, golden light, sacred feel.
Quality: 4K, cinematic quality, rich in detail.
Constraints: Consistent face, natural flowing clothes, realistic sword light effect.Case 15: Time-lapse - City Sunrise
A time-lapse style video showing a modern city changing from night to sunrise.
Camera: Fixed camera, wide-angle overhead shot, showing the entire city skyline.
Scene: City lights gradually turn off, the sky changes from deep blue to orange-red, and finally, sunlight bathes the city.
Style: Time-lapse, ultra HD, saturated colors, stunning feel.
Quality: 4K, ultra HD, extremely rich in detail.
Constraints: Stable footage, natural color transition, no flickering.Case 21: High-end Watch Advertisement
A precise Swiss mechanical watch rotates slowly on a black velvet background, showcasing its exquisite dial and gears.
Camera: Macro close-up, 360-degree orbit, occasionally switching to a side view to show thickness.
Scene: Pure black background, single point light source, metal texture shining under the light.
Style: Luxury ad aesthetic, minimalism, high contrast, exquisite feel.
Quality: 4K, ultra HD, realistic metal reflection, extremely rich in detail.
Constraints: Accurate watch proportions, smooth rotation, realistic light reflection, no deformation.Chapter 5: Advanced Techniques & Troubleshooting Guide
5.1 The Art of Multimodal Input: The Magic of @
-
Locking a Character with an Image (@image1): Provide a high-quality, front-facing, evenly lit character image. In the prompt: "A girl from @image1, walking in the park."
-
Defining Motion with First and Last Frames (@image1, @image2): Upload a screenshot of point A as @image1 and point B as @image2. Describe: "Make the character from @image1 move smoothly to the position of @image2."
-
Referencing Camera Movement with a Video (@video1): "A man standing on a mountaintop, filmed using the camera movement from @video1."
-
Driving Lip-Sync and Rhythm with Audio (@audio1): "A cartoon lion from @image1, speaking according to the audio content of @audio1, with precise lip-sync."
Advanced Combination Example:
[Prompt]: Make the astronaut from @image1 (character image), in the environment of @image2 (Mars background image), perform the running action from @video1 (a running video), while playing the tense background music from @audio1.5.2 Character Profiles for Consistency
- Create a Profile in the platform's asset library (e.g. "Agent Emily").
- Upload multi-angle reference images: front, side, 45-degree, different expressions.
- In the prompt: "Have the character 'Agent Emily' perform a mission, with facial features consistent with her profile."
5.3 Troubleshooting Checklist
| No. | Issue | Possible Cause | Solution |
|---|---|---|---|
| 1 | Flickering, Subject Deformation | Action too large or fast; drastic lighting changes | Simplify action; stabilize lighting; add constraints: stable face, no distortion, no flickering. |
| 2 | Character "Face-Swapping" or Costume Change | No character reference; lack of consistency constraints | Upload character image with @image; use character profile; add character consistency, clothing consistency. |
| 3 | Stiff, Unnatural Motion | Verb too simple or abstract; violates physics | Detail the action; upload @video reference; enable advanced physics simulation. |
| 4 | Blurry Image, Lack of Detail | No quality requirements; style too abstract | Add 4K, ultra HD, rich in detail, cinematic quality; provide style reference image. |
| 5 | Content Doesn't Match Description | Prompt too complex or contradictory | Simplify prompt; rephrase (e.g. "a 500-meter-tall skyscraper"); remove contradictions. |
| 6 | Watermarks, Unwanted Text | Training data remnants | Use negative prompts: no watermarks, no logos, no text. |
| 7 | Inaccurate Lip-Sync | Poor audio; face obstructed | Use clean vocals; use front-facing, unobstructed character image. |
Chapter 6: Conclusion - You Are the Director
Seedance 2.0 is not just a tool; it's more like a talented and tireless virtual production team. Through this guide, you have mastered the language to communicate with this team—prompt engineering.
We started by understanding the "director's mindset" of Seedance 2.0, learning its core capability of multimodal fusion. We dissected the structure of a perfect prompt and built a comprehensive "vocabulary" of camera language, lighting, style, and constraints. More importantly, through a vast case library, you have witnessed how these theories are transformed into spectacular video clips in practice.
Now, all you need to do is to constantly practice, experiment, and create. Don't be afraid to fail, because every failure is a valuable learning opportunity. Use the "Troubleshooting Checklist" to diagnose problems, optimize your prompts, and gradually establish your own creative style and methodology.
The wave of AI video generation has arrived, and you, armed with this ultimate guide, are standing at its crest. From now on, forget about complex software and expensive equipment. Your imagination is the only set; your prompt is the only script. Go create, go express, go amaze the world.
You are the director.
References
- GitHub - ZeroLu/awesome-seedance - The ultimate collection of high-fidelity Seedance 2.0 prompts and Seedance AI resources.
- Seedance 2.0 Prompts - High-Quality AI Video Generation Prompts and Examples - Atlas Cloud.
- Seedance 2.0 Detailed Tutorial: Everyone is an AI Director | MrFive on Binance Square.
- 18 Powerful Seedance 2.0 Prompts to Take Your Videos to the Next Level - Dreamina.
- Seedance 2.0 Prompt Strategy: Master This Formula to Create Cinematic... - Zhihu.
- The Complete Guide to Seedance 2.0 Prompts: From Novice to "AI Director" - CSDN Blog.