Question 1

How do I create my first video?

Accepted Answer

Simply sign up for an account, navigate to the Dashboard, and click 'Create New Video'. Choose your AI model (Veo 3, Runway, Kling, or others), enter your text prompt describing the video you want, select your preferences (aspect ratio, duration, scenes), and click Generate. Your video will be ready in minutes!

→ Ready to start? Go to New Campaign

Question 2

Do you offer a free trial or free tier?

Accepted Answer

Yes! We offer a free tier with 12 credits to help you get started. That's enough for 4 Veo 3.1 Lite videos (3 credits each). Sign up to start creating.

→ See all plans on the Pricing page

Question 3

What do I need to know before starting?

Accepted Answer

Before creating your first video, familiarise yourself with writing effective prompts. Be specific and descriptive. Start with simpler videos to understand the models, then experiment with longer durations and advanced features. Check out our template library for inspiration and examples.

→ Browse Templates for inspiration

Question 4

Which regions is ClipStudios.AI available in?

Accepted Answer

ClipStudios.AI is currently available in the European Union and United Kingdom. We're working on expanding to additional regions soon. If you're outside these regions, you can join our waitlist to be notified when we launch in your area.

Question 5

How do I write an effective prompt?

Accepted Answer

Be specific and descriptive. Include details about the scene, action, lighting, camera movement, and mood. For example, instead of 'a dog running', try 'a golden retriever running through a sunlit meadow at sunset, slow motion, cinematic lighting'. The more detail you provide, the better the result.

Question 6

What should I include in my prompt?

Accepted Answer

A good prompt typically includes: Subject (what or who is in the scene), Action (what is happening), Setting (where it takes place), Style (cinematic, documentary, animated, realistic), Lighting (natural, dramatic, soft), Camera Movement (pan, zoom, static, dolly, tracking), Mood (energetic, calm, mysterious, dramatic), and Time of Day (morning, sunset, night if relevant).

Question 7

What should I avoid in prompts?

Accepted Answer

Avoid vague descriptions, conflicting instructions, and overly complex scenes with too many elements. Keep prompts focused on one main action or subject. Also, avoid requesting specific celebrities, copyrighted characters, or content that violates our content guidelines.

Question 8

How do I prompt for Google Veo 3 (Fast & Quality)?

Accepted Answer

Veo 3 excels at cinematic descriptions and explicit camera directions. Use vivid, sensory language to paint a picture of your imaginary world.

Key techniques: Use evocative, sensory language describing light, texture, and atmosphere. Include explicit camera movements like 'pan from left to right', 'slow dolly forward', 'aerial shot with zoom out'.

Add detailed environmental context about weather, terrain, and time of day.

For Veo 3.1 with native audio, use quotes for dialogue: '"This must be the key," he murmured'. Describe sound design, including dialogue, sound effects, and ambient noise. Use colour grading terms like 'saturated colours', 'high contrast', 'warm tones'.

Example: 'Wide-angle shot of a professional chef expertly flipping vegetables in a wok with flames rising, modern restaurant kitchen with stainless steel surfaces, camera slowly dollies forward, dramatic side lighting creating strong shadows, cinematic food photography style, shallow depth of field'.

Question 9

How do I prompt for Google Veo 3.1 with audio?

Accepted Answer

For Veo 3.1's native audio generation:

Place dialogue in quotes. Specify SFX tied to visible actions and timing ('door slam, on cut'). Use concrete ambience descriptions (rain hiss, crosswalk beep). Add music role and level instructions ('duck under dialogue'). Keep each scene to 4-8 seconds with 1-2 primary actions.

Question 10

How do I prompt for Kling AI (2.1 & 2.5 Turbo)?

Accepted Answer

Kling excels at dynamic action scenes and maintains superior temporal consistency. Focus on motion and physics understanding.

Key techniques: Clear subject and action with straightforward specifics. Motion descriptions using dynamic action verbs. Physics-aware prompts describing real-world physics. Temporal markers like 'initially... then... finally'. Focal point control with 'focus on' or 'centred on'. Depth layering describing foreground, midground, and background. Camera movement specifics including speed ('slow dolly', 'rapid pan'). Use professional cinematography terms like '35mm film', 'lens flares', 'shallow depth of field'.

Optimal structure: [Shot type] of [subject] [action/movement], [environment/setting], [camera movement], [lighting/mood], [style/aesthetic].

Example: 'Medium shot of a skateboarder executing a perfect kickflip across concrete steps, urban park with graffiti walls, camera tracks from left to right following the action, golden hour lighting creating warm tones and hard shadows, cinematic street photography style, real-time motion capture quality'.

Question 11

How do I prompt for Runway Gen 3 (& Gen 4)?

Accepted Answer

Runway is an integrated creative suite with advanced camera control features, especially good at detailed object transformations and style changes within shots.

Key techniques: Pair camera control values with detailed text prompts for intentional results. Include scene context descriptions to help the model understand transformations. Add subject motion descriptions explaining how objects should move. Specify end scene descriptions showing what the final result should look like. Use style references mentioning visual inspirations (e.g., 'inspired by Blade Runner'). Include shot type precision with camera angles and movements.

Example: 'A warrior raises his sword as the camera zooms out, revealing a dragon approaching over the mountains, dramatic evening lighting with orange and purple tones, cinematic action sequence, slow-motion sword elevation'.

Use Runway when you need precise control over specific elements within a shot or want to stylise existing footage.

Question 12

How do I prompt for Wan 2.5 / Wan 2.2 A14B Turbo?

Accepted Answer

Wan excels with detailed character rendering and fine-tuned facial expressions. Use dynamic action verbs and specific character details.

Key techniques: Dynamic action verbs describing movement and behaviour. Character appearance details specifying age, clothing, key characteristics. Environmental context with time of day, weather, atmosphere. Cinematic terminology like 'dolly shot', 'bokeh', 'three-point lighting'. Sequential actions breaking complex actions into clear steps. Temporal connectors using 'then', 'next', 'afterwards' for multi-action sequences.

Example: 'Woman in professional attire entering a modern cafe, walks to counter with confident stride, orders coffee while engaging warmly with barista, cinematic lighting with soft golden window light, camera follows in smooth tracking shot, urban contemporary aesthetic'.

Question 13

How do I prompt for Kling VIDEO 3.0?

Accepted Answer

Kling VIDEO 3.0 is built for native audio, stronger subject consistency (including element references), multi-shot narratives, and flexible duration up to 15 seconds.

Multi-Shot: Turn Multi-Shot on and describe scene coverage, shot types, and dialogue — the model plans transitions. For precise control, use Custom Multi-Shot and label shots ('Shot 1 … Shot 2 …') with duration per shot.

Image-to-Video + elements: Bind a subject to keep look and voice stable; if a voice is already bound to the element, do not set tone again in the prompt.

Native audio: Assign lines clearly per character ('Character (tone): "…"'). Video 3.0 handles three or more speakers better than 2.6. Five languages (Chinese, English, Japanese, Korean, Spanish) plus dialects and accents when tagged in the prompt.

Text in frame: Ask for readable signage, captions, or product lettering when needed.

Long clips: Use time markers for beats across 3–15 seconds.

Official guide: https://kling.ai/quickstart/klingai-video-3-model-user-guide

Question 14

How do I use Kling VIDEO 2.6 Motion Control?

Accepted Answer

VIDEO 2.6 Motion Control applies motion from a reference video (upload or Motion Library) to your character image. Add a prompt to control background, scene details, and (with the right orientation mode) camera.

Workflow: Add motion to mimic → add character image (match proportions to the video for best results) → choose Character Orientation Matches Video (default) or Character Orientation Matches Image → enter your prompt → generate.

Defaults: *Matches Video* — movement, expressions, camera, and orientation follow the motion clip; the prompt fills in other scene information. *Matches Image* — movement and expressions follow the motion clip; body orientation follows the image; you can drive camera movement via prompt (e.g. zoom in/out, camera up/down, fixed).

Good inputs: Match full-body vs half-body between image and motion; use motion with broad range, moderate speed, minimal displacement; single continuous motion clip (no cuts or camera moves in the reference); 3–30s uploads, resolution within Kling's limits.

If output is shorter than your upload (complex or very fast motion), credits may still apply — see the Kling 2.6 Motion Control FAQ.

Documentation: https://kling.ai/quickstart/motion-control-user-guide

Question 15

How do I use Kling VIDEO 3.0 Motion Control?

Accepted Answer

Motion Control drives one character in your reference image using motion from an uploaded video or a pick from the motion library — you get controlled body motion and facial performance.

VIDEO 3.0 improves on 2.6 with stronger facial consistency across angles, expressions, and longer takes.

Workflow: Upload the motion source and character image → use Bind Facial Element to Enhance Facial Consistency (create or bind an element). Element binding only applies when character orientation matches video orientation.

Reference assets: The element library uses facial information only — not outfit, hair, makeup, or props. Use clear face close-ups that match the look you want. For difficult expression or head-turn accuracy, follow Kling's reference sets (front/sides, neutral/smile, etc.) or upload a short video for richer facial data.

Prompt: Describe lighting, shot type, mood, and how the performance should feel; identity and motion source are largely established by uploads and binding.

Official guide: https://kling.ai/quickstart/motion-control-user-guide

Question 16

Do different models require different prompting styles?

Accepted Answer

Yes! While all models understand natural language prompts, they respond better to certain styles:

Veo 3: Specialises in cinematic detail & realism with focus on camera directions and sensory language (50-80 words).

Kling: Specialises in action & motion physics with dynamic verbs and physics descriptions (40-70 words).

Kling VIDEO 3.0: Multi-shot and long-form (3–15s) prompts with native audio, element binding, and optional timed beats — see the Kling 3.0 FAQ and official guide.

Kling VIDEO 2.6 / 3.0 Motion Control: Prompt for scene, background, and camera (especially in *Matches Image* mode on 2.6); motion quality depends on reference video and image matching — see the Motion Control user guide.

Runway: Specialises in stylisation & editing with style references and specific changes (50-80 words).

Wan 2.5: Specialises in character rendering with character details and sequential actions (50-70 words).

Seedance: Specialises in multi-shot narratives with scene structure and transitions (100-150 words).

Question 17

Can I use reference images with my prompts?

Accepted Answer

Yes! Image-to-video generation lets you upload a reference image and describe how you want it to animate. This is great for:

• Maintaining visual consistency using the same character across multiple videos
• Animating existing artwork bringing product photos or concept art to life
• Preserving composition keeping original framing while adding motion
• Controlling first frame locking in the starting image for consistent output

Image-to-Video Prompt Structure: Subject (main focus of your image), Action (subtle movement or motion), Background (surrounding environment), Camera Movement (how the view should shift, careful not to distort the reference).

Pro Tip: For character consistency across multiple videos, generate a master character image first, then use it as a reference for each scene.

Question 18

How can I get more consistent results?

Accepted Answer

• Use specific, detailed prompts as vague prompts produce unpredictable results
• Stick to a consistent style description maintaining the same aesthetic
• Save successful prompts as templates for reuse
• Use reference images as image-to-video maintains visual consistency better than text-only
• Keep core descriptions stable and for variations only change specific elements
• Browse our template library for proven prompts
• Optimise prompt length as 30-80 words typically works best
• Use model-specific techniques leveraging each model's strengths

Question 19

What is the best prompt structure formula?

Accepted Answer

Basic Structure for Text-to-Video:
[Shot type] of [subject], [action], [setting], [lighting], [camera movement], [mood/style]

Example: 'Close-up shot of cascading water droplets, falling slowly in soft focus, against a dark background, backlit by warm amber light, slight camera drift to the left, serene and meditative mood, shallow depth of field with bokeh effects'.

For best results:
• Lead with the most important visual element
• Include 2-3 specific camera or cinematic terms
• Describe lighting explicitly (not just 'bright' but 'golden hour sunlight')
• End with overall mood or style reference
• Avoid contradictory instructions
• Stay under 80 words for focused results

Question 20

What are the pro tips for prompt engineering?

Accepted Answer

• Use industry terminology like 'dolly shot', 'shallow depth of field', 'colour graded in cool tones'
• Be specific about motion speed with 'slow pan', 'rapid zoom', 'smooth tracking'
• Describe depth mentioning foreground, midground, and background elements
• Reference inspirations like 'cinematic style inspired by Blade Runner'
• Use comparison language such as 'like a commercial', 'documentary style', 'music video aesthetic'
• Avoid negations - instead of 'not blurry', say 'sharp focus'
• Consider pacing mentioning if movement should be slow, moderate, or fast
• Be consistent saving your favourite prompt structures for reuse

Question 21

When should I use WAN 2.6 vs Kling 2.6?

Accepted Answer

Choose based on your content type:

WAN 2.6 — Cinematic, structured storytelling
• Marketing videos, product demos, educational content
• Longer durations (10-15 seconds)
• Scene-based, deliberate pacing
• Iteration-friendly workflow

Kling 2.6 — Short-form with native audio
• Quick social clips, talking-head content
• Shorter durations (5-10 seconds)
• Built-in speech and sound effects
• Fast, single-pass generation

Question 22

Which model is best for beginners?

Accepted Answer

Kling 2.6 is easier to start with:
• Shorter prompts (1,000 char limit)
• Audio included in one pass
• Quick results for experimentation

WAN 2.6 rewards more structured prompts:
• Scene-based thinking
• Iteration workflow
• Best for users comfortable with cinematography concepts

Question 23

Which model is best for marketing videos?

Accepted Answer

WAN 2.6 is optimized for marketing, education, and product storytelling:
• Extended 15-second shots for narrative depth
• Cinematic pacing and camera control
• Professional, production-grade outputs

Use Kling 2.6 for quick social clips with native audio generation.

Question 24

Which model is best for dialogue and voiceover?

Accepted Answer

Kling 2.6 and Kling VIDEO 3.0 generate native audio (speech, ambient sound, and more languages in 3.0) in a single pass. Video 3.0 adds stronger multi-character line assignment, multilingual dialogue, and dialect or accent tags.

WAN 2.6 does not generate audio — add voiceover or music in post-production.

If audio is essential, use a Kling model with native audio.

Question 25

Which model supports longer videos?

Accepted Answer

WAN 2.6 and Kling VIDEO 3.0 support up to 15 seconds (Kling 3.0 also supports flexible durations from 3–15 seconds).

WAN 2.6:
• 5 seconds → Short scene
• 10 seconds → Narrative beat
• 15 seconds → Extended shot

Kling 2.6 supports up to 10 seconds:
• 5 or 10 seconds only

Question 26

Which model is more credit-efficient?

Accepted Answer

Credit usage varies by model and settings:

Kling 2.6
• Video-only: Standard credits
• With audio: ~2× credits (due to audio synthesis)

WAN 2.6
• Standard credits (no audio option)
• More value for longer durations (15s)

Use the Credit Calculator to estimate costs before generating.

Question 27

How do I prompt for Seedance v1 Lite?

Accepted Answer

Seedance v1 Lite is optimized for fast, efficient video generation. Keep prompts simple and direct.

Key techniques:
• One subject, one action per clip for best results
• Use degree adverbs (slowly, quickly, powerfully) for motion intensity
• For camera movement: surround, aerial, zoom, pan, follow, handheld
• Focus on clear, grounded scenes over abstract concepts
• Phrase positively — negative prompts are NOT supported
• No audio support — generates silent video only

Example: 'A cyclist rides slowly down a city street in daylight. Static wide shot.'

Question 28

What is Seedance v1 Lite best for?

Accepted Answer

Seedance v1 Lite is ideal for:

• Quick iterations — Fast generation for testing ideas
• Simple clips — Single subject, single action scenes
• Speed-first workflows — When you need results quickly
• Basic motion — Straightforward camera movements

Not ideal for:
• Complex multi-shot sequences (use v1 Pro)
• Projects requiring audio (use Seedance 1.5 Pro)
• Cinematic storytelling (use v1 Pro)

Question 29

Does Seedance v1 Lite support audio?

Accepted Answer

No — Seedance v1 Lite generates video-only output with no audio.

If you need AI-generated audio, use Seedance 1.5 Pro instead, which includes native audio generation.

Question 30

Do negative prompts work with Seedance v1 Lite?

Accepted Answer

No — Seedance v1 Lite does not support negative prompts.

Phrase your prompts positively to describe exactly what you want to see.

❌ Avoid: 'A beach scene, no people'
✅ Better: 'An empty beach at sunrise, gentle waves lapping the shore'

Question 31

What camera movements does Seedance v1 Lite support?

Accepted Answer

Seedance v1 Lite supports these camera commands:

• Surround — Camera circles the subject
• Aerial — Bird's-eye or drone view
• Zoom — Zoom in or out
• Pan — Horizontal camera rotation
• Follow — Camera follows the subject
• Handheld — Natural, slightly shaky movement
• Static — Fixed camera position

Example: 'A cyclist rides slowly down a city street. Static wide shot.'

Question 32

How should I structure prompts for Seedance v1 Pro?

Accepted Answer

Seedance v1 Pro excels at cinematic, multi-shot storytelling. Structure your prompts using this formula:

Subject + Action + Scene + Camera Language + Style & Atmosphere

Not all elements are required every time, but this structure produces the best results.

Example single shot: 'A cat yawns at the camera. Close-up shot, soft natural light, shallow depth of field.'

Example with action: 'A woman picks up the glass in front of her, takes a sip, puts it down, then stands up and walks away. The camera slowly tracks her movement.'

Question 33

Does Seedance v1 Pro support multi-shot videos?

Accepted Answer

Yes — multi-shot storytelling is a key strength of Seedance v1 Pro!

Use the keyword 'lens switch' to create cuts between different shots within a single generation.

Example:
'A futuristic laboratory, wide panoramic shot. lens switch. Close-up of a glowing quantum computer emitting red light. lens switch. Low-angle close-up of a scientist reacting in shock.'

Important tips:
• Maintain subject continuity across shots (same character, consistent clothing)
• Keep style and atmosphere consistent (same lighting, colour grading)
• Each shot should describe a distinct camera angle or framing
• Use 2-4 shots for best results

Question 34

What camera movements does Seedance v1 Pro understand?

Accepted Answer

Seedance v1 Pro supports advanced camera language for cinematic control:

Movement types:
• Push / Pull — Move camera toward or away from subject
• Pan — Horizontal rotation left or right
• Track — Camera follows alongside moving subject
• Orbit — Camera circles around the subject
• Follow — Camera trails behind or ahead of subject
• Crane — Vertical camera movement (up or down)
• Zoom — Lens zoom in or out

Special shots:
• Aerial shots — Bird's-eye or drone perspectives
• Macro shots — Extreme close-ups with shallow depth
• Close-up / Wide shot / Medium shot — Standard framing
• Low-angle / High-angle — Camera positioning

Example: 'The camera slowly orbits the dancer as she spins, then pulls back to reveal the full stage.'

Question 35

Can I control visual style and aesthetics with Seedance v1 Pro?

Accepted Answer

Yes! Seedance v1 Pro has strong style control capabilities.

Art styles:
• 2D / 3D animation
• Clay, voxel, felt, illustration
• Cinematic, arthouse, retro film looks
• Anime, cartoon, realistic

Atmosphere controls:
• Lighting: soft, dramatic, neon, golden hour, moonlit
• Mood: melancholic, whimsical, tense, peaceful
• Era: 1920s noir, 80s synthwave, futuristic
• Film aesthetics: film grain, lens flares, shallow depth of field

Example: 'Clay animation style, a chubby frog relaxes in a pink massage chair while a white cat massages its shoulders. Warm lighting, whimsical tone.'

Question 36

Does Seedance v1 Pro support audio?

Accepted Answer

No — Seedance v1 Pro generates video-only output with no audio.

If you need AI-generated audio (dialogue, sound effects, ambient sounds), use Seedance 1.5 Pro instead, which includes native audio generation.

Seedance model comparison:
• v1 Lite → Fast, simple clips, no audio
• v1 Pro → Cinematic multi-shot, no audio
• 1.5 Pro → Motion-focused with AI audio

Question 37

Do negative prompts work with Seedance v1 Pro?

Accepted Answer

No — Seedance v1 Pro does not support negative prompts.

Instead of describing what you don't want, phrase your prompts positively to describe exactly what you want to see.

❌ Avoid: 'A forest scene, no people, no buildings'
✅ Better: 'A serene forest clearing with ancient oak trees, soft morning mist, untouched wilderness'

Question 38

What is Seedance v1 Pro best used for?

Accepted Answer

Seedance v1 Pro excels at:

• Cinematic storytelling — Create narrative sequences with multiple shots
• Multi-shot sequences — Use 'lens switch' for dynamic scene changes
• Stylized scenes — 2D, 3D, clay, illustration, and artistic looks
• Creative narratives — Short films, trailers, music video concepts
• High-quality 1080p visuals — Professional-grade output

Best for:
• Filmmakers and content creators
• Storyboard visualization
• Commercial and advertising concepts
• Social media content with cinematic flair

Not ideal for:
• Projects requiring audio (use Seedance 1.5 Pro instead)
• Simple, single-action clips (use v1 Lite for speed)

Question 39

What is Seedance V1.5 Pro best used for?

Accepted Answer

Seedance V1.5 Pro is best for motion-focused, cinematic videos where subject movement, camera direction, and action intensity are important.

Ideal use cases:
• Dynamic scenes with expressive motion
• Storytelling and narrative content
• Marketing videos and social content
• Projects requiring camera movement and action
• Complete video outputs with optional AI-generated audio

Key strength: Motion control with optional audio generation.

Question 40

How should I structure prompts for Seedance V1.5 Pro?

Accepted Answer

Seedance V1.5 Pro works best with motion-first prompts.

Recommended structure:
```
Subject + Motion
Background + Motion
Camera + Motion
```

Key principles:
• Focus on what moves, not static details
• The model automatically expands scene details
• Keep prompts clear, direct, and action-oriented

Example:
'A dancer leaps gracefully across the stage. The camera follows smoothly from left to right. Slow, elegant motion under warm spotlight.'

Question 41

How do Image-to-Video prompts work with Seedance V1.5 Pro?

Accepted Answer

For Image-to-Video with Seedance V1.5 Pro:

Best practices:
• Write prompts based on the content of the image
• Clearly describe the main subject and the motion you want
• Avoid describing static elements that already exist in the image
• Do not contradict the image's content or composition

Example:
'The old man wears glasses.'

The model extracts scene details from the image and animates the subject naturally.

Question 42

Does Seedance V1.5 Pro support multiple actions in one prompt?

Accepted Answer

Yes! Seedance V1.5 Pro has strong support for multiple consecutive actions.

You can list actions in sequence:
```
Subject + action 1 + action 2
Subject 1 + action + subject 2 + action
```

The model understands timing and motion continuity, expanding these actions smoothly.

Example:
'A woman picks up the glass, takes a sip, puts it down, then stands up and walks away.'

Question 43

What camera movements does Seedance V1.5 Pro support?

Accepted Answer

Seedance V1.5 Pro has strong camera language understanding. Describe camera movement using natural language:

Supported movements:
• Pan — Horizontal rotation
• Zoom — Zoom in or out
• Follow — Camera follows the subject
• Aerial — Bird's-eye or drone view
• Handheld — Natural, slightly shaky movement
• Surround — Camera circles the subject

Camera movement works for both Text-to-Video and Image-to-Video prompts.

Question 44

Can I create multi-shot videos with Seedance V1.5 Pro?

Accepted Answer

Yes! Seedance V1.5 Pro supports multi-shot sequences.

To create shot changes:
• Use the keyword 'Shot Switch' (or 'camera switch')
• Describe the connection between shots
• If the scene changes, describe the new scene clearly

Example:
'Kittens and puppies eat cat food. Shot Switch. Close-up cat food is distinct.'

Tips:
• Maintain subject continuity across shots
• Keep style consistent
• Use 2-4 shots for best results

Question 45

How do I control motion intensity and speed?

Accepted Answer

Use adverbs of degree to clearly define how strong or fast the motion should be.

Common examples:
• slowly — Gentle, measured movement
• quickly — Fast, energetic motion
• powerfully — Strong, impactful movement
• violently — Aggressive, intense motion
• wildly — Chaotic, uncontrolled movement
• with large amplitude — Exaggerated motion range

Being explicit helps the model avoid guessing motion intensity.

Question 46

Do negative prompts work with Seedance V1.5 Pro?

Accepted Answer

No — Seedance V1.5 Pro does not support negative prompts.

Instead of describing what you don't want, clearly describe what you want to happen in the scene.

❌ Avoid: 'A person walking, no running'
✅ Better: 'A person walks slowly and deliberately across the room'

Question 47

How should I prompt for Text-to-Video with Seedance V1.5 Pro?

Accepted Answer

For Text-to-Video, use a structured description:

Format:
```
Subject + Movement + Scene + Camera / Shot + Style
```

Helpful techniques:
• Detailed character descriptions (appearance, clothing, posture)
• Environmental details (natural or built spaces)
• Combining emotion with motion
• Describing atmosphere using lighting and time of day
• Camera movement and sequential actions
• Adverbs of degree for motion intensity

Example:
'A young woman in a red dress dances gracefully in a sunlit ballroom. The camera slowly orbits around her. Warm golden light, elegant atmosphere.'

Question 48

Does Seedance V1.5 Pro generate audio?

Accepted Answer

Yes! Seedance V1.5 Pro supports optional AI-generated audio, including:

• Sound effects
• Ambient audio

How it works:
• Audio can be enabled or disabled before generation
• Enabling audio consumes additional credits
• Audio is generated alongside video in one pass

Note: If you don't need audio, disable it to save credits.

Question 49

When should I use Fixed Lens vs Unfixed Lens?

Accepted Answer

Fixed Lens:
• Use for stable, static camera shots
• Best when you want the camera to stay in one position
• Ideal for dialogue scenes or focused compositions

Unfixed Lens:
• Use when your prompt includes camera movement
• Required for: pan, zoom, follow, aerial, surround, etc.
• Recommended whenever you describe camera motion

Rule of thumb: If your prompt describes camera movement, use Unfixed Lens.

Question 50

What's the difference between Seedance 2, Seedance 2 Fast and Seedance 2 Mini?

Accepted Answer

All three run the same model family and take the same prompt shape — they differ in output ceiling and cost:

• Seedance 2 — the full model: up to 15s, 480p → 4K, up to 9 reference images. Use it for finished work.
• Seedance 2 Fast — same inputs, capped at 480p/720p and 15s. Use it to lock a prompt in fewer credits before a 1080p/4K run.
• Seedance 2 Mini — the cheapest of the three, 480p/720p and 15s. Use it for rough drafts and volume tests.

Because the prompt structure is identical, a prompt tuned on Mini or Fast carries straight over to Seedance 2 — draft cheap, finish high.

Question 51

What is the basic prompt formula for Seedance 2?

Accepted Answer

For a simple shot, describe subject + action + scene + camera and stop there.

When you attach references, say what each one is for:

• Subject — 'Reference the woman in Image 1 to generate…'
• Motion or camera — 'Reference the camera movement in Video 1, not its content, to generate…'
• Voice or mood — 'Reference the voice in Audio 1 to…'
• Style or lighting — 'Refer to Image 2 for lighting and colour'

You can combine them in one prompt: 'Reference the woman in Image 1 and the lighting of Image 2, following the camera move in Video 1.' A reference the prompt never mentions gets used loosely, so bind every asset in the text.

Question 52

What is the advanced prompt formula for Seedance 2?

Accepted Answer

The model reads your prompt as two layers — space (what's in the frame) and time (how it changes) — so write it as direction, not as a caption:

```
subject + action detail + scene/environment + lighting & colour
+ camera movement + visual style + image quality + constraints
```

❌ Weak: 'A man runs nervously down the street, very cinematic.'

✅ Strong:
'Shot 1: Side shot of a street alley; the man starts running, breathing fast.
Shot 2: He knocks over a fruit stand; the camera shakes and cuts to a close-up of his frightened face.
Shot 3: He climbs a low wall and disappears; the camera pulls back slowly and holds on the empty street.
HD cinematic texture, cool tones, soft light. No subtitles, no logo, no watermark.'

Question 53

How do I structure a multi-shot Seedance 2 video?

Accepted Answer

Write a storyboard using Shot 1 / Shot 2 / Shot 3 labels, in the order events actually happen.

Organise each shot in this order:
1. Camera move or transition — 'slow push in from a wide shot', 'fixed camera', 'cut to…'
2. Subject action and expression — the key movement and how the face changes
3. Position or spatial change — where the subject is and where they go
4. Audio — sound effects, dialogue, music for that shot

Don't pin exact seconds. Timing instructions like '0–3 seconds' are unreliable and often break the generation — let the plot set the pacing.

Question 54

How do I stop the model from mixing up my characters?

Accepted Answer

Define every subject explicitly before you use it, then reuse the same label every single time.

Pattern: 'Define the [2–3 stable features] in Image 1 as [label].'

• Use stable, static features — clothing, hairstyle, build, category. Not mood or pose.
• Give each subject a unique label ('police officer', 'thief', 'Subject 1') and never swap it mid-prompt.
• For undefined subjects in simple scenes, tag them inline every time: 'Zhang San@Image 1'.
• Even when you pick assets from your library, still refer to them as Image 1 / Video 1 / Audio 1 — the model can't map a library ID to a reference on its own.

Example: 'Define the tall man in Video 1 as police officer, and the shorter man as thief. Thief runs through the market while police officer follows close behind.'

Question 55

How should I describe actions and emotions?

Accepted Answer

Be specific about the body, and quantify it. Name the part (hands, head, shoulders, back) plus range, speed and force: 'slowly raises a hand', 'quickly turns her head', 'slightly lowers his chin'.

Prefer small, continuous motion. Slow, gentle, connected movements hold up far better than sprints, big jumps and violent rolls.

Bridge one action into the next: 'uses the momentum of the turn to naturally raise a hand'.

Externalise emotion instead of naming it:
• Sadness → head lowered, shoulders trembling, fingers clutching a sleeve
• Joy → mouth rising uncontrollably, light steps, humming
• Anxiety → checking a watch, tapping the table, quick breathing
• Anger → fists clenched, jaw tight, chest heaving
• Relief → a long breath out, shoulders dropping, a faint smile

Question 56

How much camera direction should I give Seedance 2?

Accepted Answer

Use standard film terms — the model understands them directly: medium shot, close-up, wide shot, slow push in, lateral tracking, orbit, fixed shot, low angle, aerial.

Use one camera move per shot. Asking for push + pull + pan + track in the same shot is the fastest way to get an unstable, wobbling frame. If you need a second move, make it a second shot.

Question 57

How do I use quality, style and constraint words?

Accepted Answer

These three close out a good prompt and are the main defence against visual glitches:

• Image quality — 'HD', 'rich detail', 'cinematic texture', 'natural colour', 'soft light'
• Style — 'retro film', '2D Japanese anime style', 'cyberpunk cool blue-purple grade', 'fresh Japanese look'
• Constraints — 'no subtitles', 'avoid generating any text or subtitles', 'do not generate a logo', 'do not generate a watermark'

Unlike the older Seedance v1 models, Seedance 2 does respond to constraint wording — leaving it out is what lets stray captions and platform watermarks through.

Question 58

What are the bracket symbols in Seedance 2 prompts?

Accepted Answer

Seedance 2 generates audio natively, and brackets tell it which kind of sound a line is: • ( ) → music — '(fast-paced rock plays in the background)' • < > → sound effect — '' • { } → dialogue — 'she says {Hello, world}' • 【】 → subtitle — '【Chapter One: Departure】' Keep dialogue in one language. Mixing languages in a line (proper nouns aside) garbles the voice track. For a less common language, mark it: 'says in Japanese {こんにちは}'.

Question 59

How many reference images and videos should I attach?

Accepted Answer

4–5 assets is the sweet spot, filling four roles:

1. Character anchor — 1–2 images that lock the face and outfit
2. Scene setter — 1 image for the environment and tone
3. Camera reference — 1 video for shot language and action rhythm
4. Rhythm and voice — 1 audio clip for mood and timbre

Don't fill the slot limit just because you can. Too many references and the model can't work out which one takes priority — you get style conflicts, blurred subjects and results that drift from the brief.

Upload in the order you want to reference them, and put the assets that need the most exact match first.

Question 60

Why does my character's face change halfway through the video?

Accepted Answer

That's ID drift, and it almost always traces to a weak face reference — either one combined image carrying face, pose, outfit and details at once, or a face that occupies too little of the frame for the model to weight it properly.

Fix it:
1. Add a dedicated headshot — head only, neutral expression, minimal shoulders/neck/background — alongside your full-body shot.
2. Split the roles in the prompt: 'Subject 1 — facial features from Image 1 (headshot), makeup and styling from Image 2 (full body)'.
3. Put the headshot first, since the earliest assets carry the most reference weight.

Avoid multi-view character sheets. Several angles of one person often read as several different people, which makes drift worse rather than better.

Question 61

Why do two identical characters appear in my video?

Accepted Answer

The 'twin' problem shows up when the frame is crowded and the subjects aren't pinned down. To reduce it:

1. Bind every character to its reference — 'Zhang San (Image 1) throws the passbook to Li Si (Image 2)', in a consistent format.
2. Add a global constraint at the end — 'Throughout the video, no two characters may share identical appearance, clothing and accessories. Do not generate duplicate characters or twins. Keep only one of each character in frame.'
3. Use single-person reference photos, not three-view or multi-view sheets.
4. Trim the prompt. Pasting a whole script confuses the model — cut anything the shot doesn't need.

Also keep the cast to four or fewer reference people. Past that, expect the wrong headcount. For a bigger cast, generate grouped images of ≤4 people each, then use those images as references for the video.

Question 62

Why does my video have subtitles, a logo or a watermark I didn't ask for?

Accepted Answer

Stray text and platform watermarks leak in from the training data and from text visible in your own reference assets.

Reduce it:
1. Add explicit constraints: 'keep it subtitle-free', 'avoid generating any text or subtitles', 'do not generate a logo', 'do not generate a watermark'.
2. Clean text out of your reference images or clips before uploading them.
3. Generate landscape where you can — subtitles appear noticeably less often in landscape than portrait — then crop to portrait afterwards.

No method removes it 100% of the time; these lower the odds meaningfully.

Question 63

Why did my anime prompt come out looking live-action?

Accepted Answer

That's style drift — usually a realistic reference image pulling the output back toward photorealism while the prompt says nothing about style.

Fix it: state the style explicitly — '2D Japanese anime style', '3D Chinese-style comic', 'clay animation'. For tight control, convert the reference image into the target style first, then generate the video from the restyled image.

Question 64

Can I extend a Seedance 2 clip or make it longer than 15 seconds?

Accepted Answer

Not as a true extension in ClipStudios today. Every run is a fresh generation of up to 15 seconds, and a reference video guides the new shot rather than being continued in place.

To build something longer:

• Generate each beat separately and stitch them in your editor
• Use frames mode with the last frame of one clip as the first frame of the next, so the join lines up
• Keep resolution and aspect ratio identical across the generations

For a single continuous take beyond 15 seconds, use Seedance 2.5, which generates up to 30 seconds in one pass.

Frequently Asked Questions

Getting Started

Prompting Tips

Choosing the Right Model

Seedance v1 Lite

Seedance v1 Pro

Seedance V1.5 Pro

Seedance 2 (2.0, Fast & Mini)

Seedance 2.5

Kling 2.1

Kling 2.5

Kling 2.6

Kling VIDEO 2.6 Motion Control

Kling VIDEO 3.0

Kling VIDEO 3.0 Motion Control

WAN 2.2

🎭 Emotional Video Creation with WAN 2.2

WAN 2.5

WAN 2.6

WAN 2.7

WAN 3.0

Runway Gen-3 Alpha

Veo 3.1

Veo 3.1 Lite

Quality & Capabilities

Pricing & Subscription

Commercial & Licensing

Security & Privacy

Technical & Support

Video Tutorials

Still have questions?