Seedance 2.5 Flash

ByteDance Seedance 2.5 Flash, the cheaper tier of the Seedance 2.5 card (set Model to Seedance 2.5 Flash). Everything Seedance 2.5 does at about 35 percent less per second (a run with a reference clip holds credits for the longest output it could make until it finishes), at 480p or 720p only (no 1080p), with no draft mode and at most 5 reference videos per run. Not independently browsable.

Last updated View as MarkdownAgent setup

Seedance 2.5 Flash

Node type: video:seedance_2_5_flash
Category: video
Hidden in sidebar: yes (auto-selected by routing)

Description

ByteDance Seedance 2.5 Flash, the cheaper tier of the Seedance 2.5 card (set Model to Seedance 2.5 Flash). Everything Seedance 2.5 does at about 35 percent less per second (a run with a reference clip holds credits for the longest output it could make until it finishes), at 480p or 720p only (no 1080p), with no draft mode and at most 5 reference videos per run. Not independently browsable.

Pricing

  • Cost: ~18 credits per second
  • Per-second rate: The figure is the rate for a 5-second clip. Shorter clips can cost a bit more per second. You're always charged for the real length.

Variant family

  • Family: bytedance_seedance
  • Is primary: no
  • All variants: video:seedance_2_5_flash

Canvas ports

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

ID Label Details
prompt Prompt TEXT (required) · multiline
first_frame_url Start Frame IMAGE. The video opens exactly on this image. Use it when you have already composed the shot (framing, background, camera position) and want it kept. Costs one of the registered materials (up to 8 per run, 5 on the older inline route). Pair it with an End Frame for a start-to-end shot. It cannot be combined with any reference media (Reference Images, Reference Audio, Audio): Seedance runs a start/end-frame task OR a reference task, never both (the mix is refused before you are charged). With a Reference Video wired, this image is sent as one more reference image instead, so the video no longer opens on it. A Start Frame shot cannot take your own voice: with Generate Audio on, the model makes its sound.
image1 Reference Image 1 IMAGE. An identity and style guide, not a framing guide. The model may reframe freely, so a carefully composed shot wired here comes back as a DIFFERENT shot. If you want the video to open on this exact image, use Start Frame instead.
video_url Reference Video VIDEO · list
audio_url Reference Audio AUDIO · list. Audio the model listens to; this track never plays in the result. In an edit (a Reference Video wired and a prompt that changes it, like "replace the man in @Video1"), wire the audio of that same clip here, turn Generate Audio off, and add this line to the prompt: "Match the framing of @Video1 in every frame: same face size and position, same distance from the camera. Never move closer or crop tighter." Without it, the audio can pull a wide shot into a close-up. The mouth then follows the timing and big shapes of this track, close but not frame perfect, and the clip comes back silent. Check its length against the clip (short clips have come back about 6 percent fast), retime it in the Video Editor if it is off, then lay the audio back in. Keep Generate Audio off when the track is a copyrighted song, because with it on Seedance can refuse the generated audio for copyright. With Generate Audio on, Seedance re-voices the track (timbre, timing and words), so the lips will not match it exactly. Without a Reference Video, it cannot be combined with a Start Frame (the mix is refused before you are charged). mp3 or wav only.
image_url Reference Image (from Seedance 2.0) IMAGE. An image arriving from the Seedance 2.0 card. It is used as a reference, not as an opening frame: the model may reframe freely. To make the video open on an exact image, wire the Start Frame port above.
end_image_url End Frame IMAGE. Optional last frame. Beside a Start Frame the video ends exactly on this image (and, like the Start Frame, it cannot be combined with reference media). With no Start Frame it is treated as one more reference image, so it guides the ending rather than pinning it.
reference_image_urls Reference Images IMAGE · list. Identity and style guides. The model may reframe freely, so these steer WHO and WHAT is in the shot, not how it is framed; to pin the framing use Start Frame. Start Frame, End Frame and any reference VIDEO all draw on the same budget of registered materials per run (up to 8, or 5 on the older inline route).
audio_urls Audio AUDIO · list. Optional reference audio. Shares the reference-audio slot with Reference Audio, and wiring both concatenates them: up to 2 tracks and 30 seconds between the two. It behaves exactly like Reference Audio (see that port for lip sync in an edit) and never plays in the result. Without a Reference Video, it cannot be combined with a Start Frame.

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, 9:16, adaptive · default: "16:9"
duration Duration TEXT · options: auto, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30 · default: "5"
resolution Resolution TEXT · options: 480p, 720p · default: "720p". Seedance 2.5 Flash makes 480p or 720p only. For 1080p or 4K, set Model back to Seedance 2.5.
generate_audio Generate Audio BOOLEAN · default: true. On: the clip comes back with sound the model generates, guided by any Reference Audio. Off: the clip comes back SILENT, and it never keeps your own track. For lip sync in an edit, turn this off, wire the audio of the clip being edited into Reference Audio, and add this line to the prompt: "Match the framing of @Video1 in every frame: same face size and position, same distance from the camera. Never move closer or crop tighter." The mouth then follows the timing and big shapes of that track, close but not frame perfect; check the take length against the clip and retime it in the Video Editor if it is off, then lay the audio back in. Keep it off when that audio is a copyrighted song, because with it on Seedance can refuse the generated audio for copyright. Outside an edit, with this on, Seedance re-voices any Reference Audio, so the lips will not match your own track exactly.
watermark Watermark BOOLEAN · default: false
contentFilter Content filter BOOLEAN · default: true. On by default. Seedance 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 Seedance 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, 9:16, auto. 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.
draft Draft BOOLEAN · default: false. Seedance 2.5 Flash has no draft mode; a run with this on is refused before you are charged.

Outputs

ID Label Type
video Generated Video VIDEO

Seedance 2.5 Flash

Seedance 2.5 Flash is a setting on the Seedance 2.5 card, not a separate node. Set Model to Seedance 2.5 Flash in the sidebar. Over the API or MCP, set quality: "flash" in the Seedance 2.5 node's config.

Flash costs about 35 percent less per second than Seedance 2.5. Credits with no reference video, by length:

Length 480p, 2.5 480p, Flash 720p, 2.5 720p, Flash
5 s 75 50 135 90
10 s 150 100 270 180

A run with a reference video holds credits for the longest clip it could make until it finishes, then settles on what the run used, so the difference comes back. With Duration on Auto that is 30 seconds. Duration 5 is the default and holds like Auto, 30 seconds. Any other length holds that length or the reference clip's, whichever is longer. That hold can be larger than Seedance 2.5's for the same short clip: at 720p with a 5 second reference clip and Duration on Auto, Flash holds 860 credits where Seedance 2.5 holds 611, and you need that balance to start the run.

What Flash does: everything the card does (a prompt, a Start Frame and End Frame, reference images, video and audio, native audio, 4 to 30 seconds or Auto), at 480p or 720p.

What it does not do: 1080p, 4k, draft mode, or more than 5 reference videos. Flash takes at most 10 reference images, 5 videos and 5 audio tracks; the card already takes at most 8 images and videos together and 2 audio tracks, so the video cap is the one you can reach. Each of those is refused before you are charged; set Model back to Seedance 2.5 for those shots.


Auto-generated from the Wireflow node registry.

For AI agents: the full documentation index is at /llms.txt, and most docs pages are available as Markdown by adding .md to the URL.

© 2026 Wireflow. All rights reserved.