Seedance 2.5

Node type: video:seedance_2_5
Category: video

Description

ByteDance Seedance 2.5. Generates video from a prompt alone, or from reference material: up to 4 reference images, a reference video, and a reference audio track. Native audio generation, so the clip comes back with sound. Up to 30 seconds of output and up to 30 seconds of combined reference footage, at 480p or 720p (there is no 1080p tier on this model). 🔴 TASK-CLASSIFICATION TRAP, read this before writing a prompt: the model picks its task from the PROMPT TEXT. A prompt containing editing or extension trigger words ("edit", "add", "remove", "replace", "extend", "continue") is treated as video-editing or video-extension rather than generation, and that mode REQUIRES Aspect Ratio "adaptive" (and, for editing, duration -1). Sending a normal aspect ratio or duration with such a prompt fails after submission with InvalidParameter.TaskTypeConstraint. Wireflow documents this and does not enforce it: phrase generation prompts without those verbs, or set Aspect Ratio to adaptive.

Pricing

• Cost: ~26 credits per second
• Notes: RETAIL prices (marginMultiplier 1.0 bypasses progressive tiers). Derivation, same hw24/1024 token formula as 2.0: 720p 16:9 (1280x720) = 21,600 tokens/s, x $10.7/M = $0.2311/s true cost, x1.13 margin = $0.26/s retail. 480p 16:9 (854x480) = 9,608 tokens/s, x $10.7/M = $0.1028/s true cost, x1.4 margin = $0.1439, rounded up to $0.15/s retail. Cross-check against BytePlus published examples: a 5s 720p clip with no reference video = 21,600 x 5 x $10.7/M = $1.156, matching their quoted $1.16. With-video-input runs bill a cheaper $6.4/M but ALSO bill the reference footage tokens: covered via tokenPricing, pre-charge adds a 30s allowance and settlement trues down to usage.total_tokens (never up). Margins copied from the 2.0 directive and NOT yet confirmed for 2.5. refVideoTokensPerSecond is 21,600, a CEILING (2.5s own 720p output token density), not a measurement: settlement only trues down, so the reserve must not come in short. One live 2.5 reference-video run must replace it before the tier is featured.

Variant family

• Family: bytedance_seedance
• Is primary: no
• All variants: video:seedance_2_5

Canvas ports

These appear as port handles on the left side of the node.

ID	Label	Details
prompt	Prompt	TEXT (required) · multiline
image1	Reference Image 1	IMAGE
video_url	Reference Video	VIDEO
audio_url	Reference Audio	AUDIO
image_url	Start Frame	IMAGE — Optional first frame. The generated clip opens on this image.
end_image_url	End Frame	IMAGE — Optional last frame. Treated as another reference image, so it guides the ending rather than pinning it exactly.
reference_image_urls	Reference Images	IMAGE — Up to 4 images describing the subject, style or scene. Start Frame and End Frame count toward the same 4.
audio_urls	Audio	AUDIO — Optional reference audio. Same slot as Reference Audio; wire either one.

Sidebar config

These render as form fields in the right-side config panel when the node is selected.

ID	Label	Details
ratio	Aspect Ratio	TEXT · options: 21:9, 16:9, 4:3, 1:1, 3:4, +2
duration	Duration	TEXT · options: 4, 5, 6, 7, 8, +22
resolution	Resolution	TEXT · options: 480p, 720p
generate_audio	Generate Audio	BOOLEAN
watermark	Watermark	BOOLEAN
aspect_ratio	Aspect Ratio override (set by Seedance 2.0)	TEXT · options: 21:9, 16:9, 4:3, 1:1, 3:4, +2 — Leave this empty. It is filled in automatically when the Seedance 2.0 node routes a run into this model, and while it has a value it OVERRIDES the Aspect Ratio field above ("auto" lets the model choose from your reference media). To change the shape of your video, use Aspect Ratio, not this.

Outputs

ID	Label	Type
video	Generated Video	VIDEO

---

Auto-generated from the Wireflow node registry.