When to use
- Marketing clips
- Product showcases
- Social vertical video
text_prompt → generate_video, or upload_image → generate_video.
Inputs
Connect a text prompt and/or a first-frame image to generate video. Several inputs appear only when the selected model supports them.text
The prompt describing the video. Optional. Accepts an array and supports loop mode.
Inline prompt. You don’t need a separate Text Prompt node for a fixed prompt —
type it directly on the node. The inline field appears whenever the prompt input is
unconnected and disappears once you wire a text input in. Connect a Text Prompt node
only when you want to drive the prompt from upstream or expose it as an app input.
Type
@ to mention a connected input by name — text inputs insert their content into
the prompt, media inputs insert their name.image
An image used as the opening frame. Optional. Accepts an array and supports loop mode. Takes one connection: wiring a second node in replaces the first.
Drawn annotations. You can draw on a still in the editor and save it with Commit annotations. When that version is the first frame, the last frame, or a reference image, the node adds one sentence to the prompt telling the model what the marks mean, for example: “Follow the drawn annotations on the first frame: a red arrow pointing at the right.” To set a motion path, draw an arrow along the way the subject or camera should move. Images without annotations leave the prompt unchanged.
brand
Brand context applied to the output. Optional, accepts multiple.
product
Product context from a Product from Website node. Product facts are injected into the prompt, and the product’s photos and logo are used as reference images. Optional, accepts multiple.
audio
One or more audio tracks. Available only when the selected model supports audio sources. Accepts multiple, up to the model’s limit, which the config panel shows (Seedance 2.5 takes up to 10). On Seedance, audio needs at least one reference image or video alongside it.
image
An image used as the closing frame. Available only when the selected model supports end frames. Takes one connection: wiring a second node in replaces the first.
video
Videos used as references. Available only when the selected model supports video sources. Accepts multiple, up to the model’s limit (Seedance 2.5 takes up to 10).
image
Images used as references. Available only when the selected model supports image sources. Accepts multiple, up to the model’s limit (Seedance 2.5 takes up to 30). Brand, product, and context images count toward the same limit, after the images you wire in here.
any
Anything to include as additional context: scraper Results, text descriptions, or mixed bundles. Contents are described to the model in the prompt. Images, videos, and audio in the bundle are passed as references where the selected model supports them. Optional, accepts multiple.
element
Subject reference slots, shown when you set element slots above zero. On models without subject-reference support, such as Seedance 2.5, element images are sent as reference images.
The node warns you, without stopping the run, when you connect more reference images, videos, or audio than the selected model takes (the extras are dropped), when Seedance gets audio with no image or video, and when a first or last frame will be sent as an ordinary reference. On Seedance 2.x, that happens whenever references are connected, because the model has no separate first-frame setting in reference mode.
Outputs
video
The generated video.
Configuration
number
default:"1"
How many video variations to generate (1 to 10).
string
default:"medium"
Amount of motion in the video. One of
low, medium, or high.number
default:"0"
Number of subject-reference element slots to expose. Synced to the selected model’s capability.
number | "auto"
Clip length in seconds. The slider snaps to the lengths the selected model supports. Seedance 2.0, Seedance 2.5, and FLUX 3 add an Auto notch (their default): the model picks the length from your prompt, up to 15 seconds on Seedance 2.0, 30 on Seedance 2.5, and 20 on FLUX 3. With Auto, credits are held at that ceiling when the run starts and then settle to the delivered clip’s real length.
This node runs on a selectable model and consumes credits. Additional model-specific options (such as duration, aspect ratio, resolution, and audio generation) are available depending on the model you choose. Pick the model in the node’s config panel. See Video models for the options, and Model settings for what each setting does.