Introduction

Seedance 2.0 is a third-party model available in Runway that generates video from text prompts, reference images, audio, and video inputs. The model supports director-level control over camera movement, lighting, and character performance, and produces audio-visual output with synchronized sound — all within a unified multimodal architecture—meaning that you set the role of your inputs. 

This article covers how to access the tool, input best practices, available settings, and what to expect from your generations.

Notice: Seedance 2.0 is likely to moderate inputs that depict realistic humans. For best results, use stylized characters or avoid inputs that depict the faces of real or photorealistic people.

Spec information

Step 1 — Selecting the inputs

Begin by navigating to the web app. There are two ways to navigate to the Seedance 2.0 model:

• Search for it beneath the What do you want to create? bar 
• Select Video, then select Seedance 2.0 from the model dropdown

Choosing the creation mode

Seedance 2.0 offers three creation modes, References, Start/End frames and Text to Video:

• References — Best for blending elements from multiple images or videos into a single generation. Use when you want granular control over what gets pulled from each input.
• Start / End frames — Best for traditional image-to-video (start frame only) or keyframe control (start and end frames). Use when you need precise control over how a shot begins or ends.
• Text to video — Best for generating without any image or video inputs. Also a good option when working with realistic human subjects, as it tends to avoid the moderation constraints that can come with image-based inputs.

Step 2 — Writing the prompt

Your prompt will vary depending on both your creation mode and what you're trying to accomplish. In any creation mode or model, we recommend using positive, unambiguous, and outcome-focused language.

Using the reference inputs

When using Reference mode, write your prompt as if you're describing the scene or sequence you want to appear, or the changes you want made, as well as how to use each provided input.

References are flexible by design. Whether you're working with images or videos, you control what gets pulled from each input and how it's used. 

• With images, you might use one reference for a subject and another for a background, or instruct the model to use a frame at a specific point in the video. 
• With videos, you can preserve the motion while changing the style, or keep the structure while swapping out characters entirely.

Below are examples of different use cases that use a combination of image and video references, their prompt, and the results of the generation.

Use case	Prompt	Input(s)	Result
Generating multi-shot video from image(s)	multishot video. the woman realizes that she forgot there was a test that day. watercolor animation style
Generating single-shot video from image(s)	use Image 1 as the starting frame for a single, continuous shot in freeze time. the camera dramatically weaves through the completely frozen scene
Determining placement of a reference image	use Image 1 as the first frame. the man slyly smiles and says "well, i guess i'll catch you later" before leaping out the airplane. a parachute deploys while he's midair.
Swapping a character or object in a video	replace the male knight in Video 1 with a woman. fiery red hair.
Restyling an existing video	relight Video 1 to dusk with a purple sky. colorize Video 1
Animating a storyboard	use Image 1 as a storyboard to guide the scenes
Apply camera motion to a new scene	apply the camera motion from Video 1 to the scene in Image 1

Step 3 — Generating the Video

Click the Generate button to start the generation. The total processing time will depend on your inputs and selected output duration, but typically takes been 2-10 minutes to complete once started.

Troubleshooting

In the event of generation errors, review the details below to learn more about troubleshooting: