Seedance 2.5

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.

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.

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
contentFilter Content filter BOOLEAN · default true. On by default. BytePlus screens your prompt, your reference material and the finished clip, so a run can fail after it has been submitted when something gets flagged (copyright is the common one). Turning this off asks BytePlus to skip that safety and IP moderation. It costs 10 percent more per run, and you are responsible for holding the rights to whatever you generate.
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.

Content filter, and what turning it off costs

BytePlus moderates a Seedance run in three separate places: your prompt, the reference images and video you register, and the FINISHED CLIP. The last one is the surprising one. A run can pass the first two checks, generate, and then fail after submission with a message like "the request failed because the output video may be related to copyright restrictions". Nothing you can see on the canvas predicts it.

Content filter is on by default and that is the behaviour every existing board already has. Switch it off and Wireflow sends content_filter: false, which asks BytePlus to skip that safety and IP moderation.

Two things to know before you do:

  • It costs 10 percent more. BytePlus bills the run at a higher rate, so the quote on the node, the credit gate before the run, the charge and the final settlement all go up by 10 percent. The number on the card is the number you pay, either way.
  • The rights are yours to hold. Skipping the filter does not give you a licence to anything. You are responsible for having the rights to the material you feed in and to whatever comes out.

Leave it on unless a run has actually been refused and you know the footage is yours to use.

Outputs

ID Label Type
video Generated Video VIDEO

Auto-generated from the Wireflow node registry.

© 2026 Wireflow. All rights reserved.

Seedance 2.5 | Wireflow Docs