How to Use Seedance 2.0

Learn how to use Seedance 2.0 in CawCut for multimodal video generation, complex motion, camera control, and synchronized audio.

Updated Nov 18, 2026 · 7 min read

Give Seedance 2.0 more than a sentence and a subject: guide it with text, images, video, or audio to shape motion, camera work, and sound together. In CawCut, Seedance 2.0 is built into the [Video Generation] node and supports Text, Image References, Frames, and Omni References. It can create multi-shot audio-video clips up to 15 seconds, handle complex interactions, and generate 4K video.

1. Seedance 2.0 Strengths

StrengthApplication
Multimodal reference control: Uses text, image, video, and audio inputs to guide composition, movement, camera, visual effects, and sound.Create a scene that combines a specific character or product appearance with a chosen action, camera move, visual style, and sound direction.
Complex, physically plausible motion: Designed to improve motion stability, physical accuracy, and interactions between subjects.Stage action with clear contact and cause-and-effect, such as a hand picking up an object, two people passing a prop, or a team performing a coordinated movement.
Multi-shot audio-video generation: Produces a connected sequence of shots and synchronized audio in a clip up to 15 seconds.Tell a compact story with a setup, a change, and a payoff, or build a short commercial with a clear visual progression.
Camera and performance direction: Responds to directions for camera movement, lighting, shadows, and subject performance.Create a cinematic reveal, follow an action through a space, or direct a performance with a specific pace and framing.
4K output: Supports high-resolution video output for detailed final delivery.Produce a polished clip for larger screens, high-detail product presentation, or a 4K finishing workflow.

2. Choose a Mode and Inputs

Select the mode in the Video Generation composer according to the references that should shape the result:

ModeUse it forInputs
[Text]Creating a scene from a written brief.A prompt describing the subject, action, setting, camera, and sound.
[Image References]Animating a still or guiding a subject's visual identity and scene composition.Images or character references, plus a direction for the movement.
[Frames]Defining a specific opening and optional final composition.A first-frame image, optional last-frame image, and a description of the transition.
[Omni References]Combining multiple reference types to direct one generation.Images, character references, video, audio, and a prompt that assigns each reference a role.

Use Text when your idea is best explained as a scene. Add references when a still image, performance, camera move, or audio cue matters to the final clip. For the full mode guide, see Video Generation.

3. Prepare References and Prompts

Seedance 2.0 works with mixed media, so make the role of each asset explicit:

  • Image or character reference: Name the subject details to preserve, such as clothing, silhouette, color, product shape, or location.
  • Video reference: State whether it should guide movement, camera behavior, framing, pace, or scene structure.
  • Audio reference: Say whether it should guide dialogue, sound effects, music, rhythm, or atmosphere.
  • Storyboard image: Explain whether the panels define shot order, composition, visual style, or copy. See Image Storyboard for planning a sequence.

For an effective video prompt:

  1. Describe the main subject and action first.
  2. Break a multi-shot clip into a short sequence of ordered beats.
  3. Give the camera one coherent move per beat, such as tracking, panning, pushing in, or holding still.
  4. Specify key interactions, including who or what makes contact and what happens next.
  5. Add lighting, visual style, dialogue, and sound direction only where they help communicate the scene.
  6. State which details should stay consistent and which details may change.

Seedance 2.0 is suited to complex motion, but a prompt can still become hard to follow if it asks for many unrelated events at once. Build the sequence around one central action and make transitions understandable.

4. CawCut Workflow and Settings

  1. Open a workflow and add a [Video Generation] node.
  2. Choose [Text], [Image References], [Frames], or [Omni References].
  3. Select Seedance 2.0 from the model menu.
  4. Add the references for the selected mode and write a prompt that explains the action and each reference's role.
  5. Set duration, aspect ratio, resolution, and audio options. Choose [4K] when you need the model's high-resolution output for a detailed final clip.
  6. Review the credit estimate, run the node, and watch the complete clip for motion, continuity, sound, and timing.

You can connect a storyboard or other workflow outputs to the Video Generation node. For help planning or connecting a multi-step workflow, see Understanding Nodes. If you combine generated clips afterward, see Video and Transition Nodes.

5. Limitations and Refinement

Seedance 2.0 can still produce unstable small details, complex multi-subject interactions, or imperfect physical movement. Audio can also contain distortion, and detailed instructions do not guarantee exact timing or text. Treat references as guidance rather than a frame-perfect lock.

If you noticeTry this
An interaction looks physically unclearDescribe the starting position, the contact or movement, and the resulting position as separate events.
The camera or subject moves differently than intendedSimplify the direction to one main camera move and one main subject action. Add a video reference for movement or camera guidance.
The character, product, or setting driftsUse a clearer reference and state which visual features must remain consistent. Remove references that suggest a conflicting design.
Audio does not match the actionSeparate dialogue, ambience, sound effects, and music in the prompt. State when each should occur.
The clip feels too crowdedReduce the number of story beats and prioritize the key interaction or visual reveal.

Change one part of the prompt or reference set at a time so you can tell which revision helped. For model alternatives and related usage notes, see Supported Models. Check Credits before generating higher-resolution or longer clips.

Related articles

Frequently Asked Questions

Learn how to use Seedance 2.0 in CawCut for multimodal video generation, complex motion, camera control, and synchronized audio.

Seedance 2.0 is designed for multimodal video creation, complex motion, camera and performance direction, multi-shot storytelling, and audio-video generation.

Seedance 2.0 supports Text, Image References, Frames, and Omni References in the Video Generation node.

Yes. Use Omni References to combine video or audio guidance with images, character references, and a prompt. Explain what each reference should contribute.

Seedance 2.0 supports 4K output in CawCut. Choose 4K from the resolution setting when preparing a high-detail final clip.

The model can still vary small details, complex interactions, and exact timing, and generated audio may occasionally distort. Keep the shot direction focused and review the full result.

Describe the action in order, identify which subject moves, and give the camera one clear direction. Use a reference clip when you want to guide movement or camera language.