Genvizu

AI creations made with Seedance 2.5

ByteDance

Seedance 2.5 is ByteDance's long-form video model on Genvizu. One run returns a single clip of 5 to 30 seconds, selectable in whole seconds, at 480p or 720p — there is no 1080p or 4K tier here. It generates audio by default and takes up to nine reference images, three reference clips totalling 30 seconds and three reference audio tracks, across seven aspect ratios including an adaptive option.

Verified on 2026-09-10

Start free · no credit card

Best for

  • Clips that have to run past 15 seconds — the duration range reaches 30 seconds, twice what the Seedance 2.0 tiers allow.
  • Continuing or restyling a longer piece of footage: reference video is accepted up to 30 seconds combined, against 15 in Seedance 2.0.
  • Multimodal briefs in one run: up to 9 reference images, 3 reference videos and 3 reference audio tracks.
  • Letting the model keep the framing of a reference instead of forcing a crop, by choosing the adaptive aspect ratio.
  • First-and-last-frame control, by putting the two stills first among the references and naming them in the prompt as @Image1 and @Image2.

Not for

  • Deliveries that must be 1080p or 4K. The tiers here stop at 720p; Seedance 2.0 Pro is the one with 1080p.
  • Clips shorter than five seconds — five is both the shortest selectable duration and the minimum billed length.
  • Multi-shot narratives from one prompt. This is a single-shot model; storyboard the cuts and generate them one at a time.
  • A guaranteed output frame size when you pick adaptive, since the ratio then follows the reference material.

Specifications and limits

FieldValue
ProviderByteDance, as Seedance 2.5 (routed through Genvizu)
InputText prompt; optional reference images, reference videos and reference audio
Output1 clip per run, with generated audio unless you switch it off
Resolutions480p / 720p (480p is the default; there is no 1080p or 4K tier)
Duration5–30 seconds, any whole second (5 is the default)
Aspect ratios16:9, 4:3, 1:1, 3:4, 9:16, 21:9, plus adaptive
Reference imagesUp to 9 per run
Reference videosUp to 3 clips, 30 seconds combined
Reference audioUp to 3 clips, alongside a reference image or video
BillingPer second of output, with a five-second minimum (tiers on the pricing page)
Billing with a reference videoA lower per-second rate, charged on input seconds plus output seconds
Plan requirementNone; available on every plan including Free
Failed runsRefunded automatically

Prompt examples

A 20-second single-take scene

Prompt

A ceramicist finishing a bowl on a wheel in a sunlit studio, 20 seconds, one continuous take.
Camera: begins wide on the room, drifts slowly in over the first half, settles into a close shot of the hands and the rim by the end.
Action: the wheel keeps turning throughout, water is added once around the midpoint, the rim thins gradually and never jumps in shape.
Audio: the hum of the wheel, wet clay, one distant street sound. No music, no voice-over.
Look: soft window light from camera left, dust in the beam, muted earth palette.

Long durations drift when the prompt describes only a moment. Give the clip a beginning, a middle and an end so the model has something to fill twenty seconds with.

Extending a longer piece of footage

Prompt

Continue the motion from the reference video: the train keeps moving through the valley away from camera, the frame holds.
Match the grade, the grain and the overcast light of the reference exactly; do not re-frame or zoom.
Audio: rail noise and wind only, no music.
Duration: 8 seconds.

Reference video is accepted up to 30 seconds combined here. Billing then counts input seconds as well as output seconds, so trim the reference to the part that actually matters.

Adaptive ratio from a reference still

Prompt

Use @Image1 as the first frame. Keep its framing and crop.
The subject lowers the camera from her eye, turns towards the window, and holds still as the light shifts.
Camera: locked off, no zoom, no handheld drift.
Audio: room tone and one shutter click only.
Duration: 10 seconds.

With the adaptive ratio the output follows the reference instead of being cropped into a fixed frame. Pick a fixed ratio instead whenever the delivery has a required frame size.

How it compares

Seedance 2.0 Pro

The tier with 1080p. Clips run 4 to 15 seconds, reference video is capped at 15 seconds combined, and there is no adaptive ratio. Pick it when the delivery needs 1080p or a clip under five seconds.

Seedance 2.0 Fast

The family’s draft tier: 480p or 720p, 4 to 15 seconds, reference video capped at 15 seconds combined. Pick it for blocking out a shot before committing to a long take here.

Examples from the gallery

Common failures and fixes

There is no 1080p or 4K option in the resolution picker.

This model tops out at 720p. Deliver at 720p, or move the run to Seedance 2.0 Pro, which has the 1080p tier.

A clip longer than 30 seconds cannot be requested.

30 seconds is the upper bound. Storyboard the sequence into separate shots and join them afterwards — the canvas has a transition step for exactly that.

A reference video is rejected, or the billed seconds are higher than the clip length.

Up to three clips totalling 30 seconds are accepted, and their real duration is read on upload. A run with reference video is billed on input seconds plus output seconds, so trim the reference before attaching it.

The audio track is unwanted or wrong.

Audio is generated by default. Either switch it off, or describe what should be heard: room tone, specific effects, and an explicit "no music, no voice-over" when you plan to add your own.

The output frame size is not what the delivery needs.

The adaptive ratio follows the reference material rather than a fixed frame. Choose one of the six fixed ratios when the frame size is part of the spec.

A long clip loses the subject or cuts to a different scene halfway.

That usually comes from a prompt written for a single moment, or one that describes two scenes. Keep one subject, one location and one camera move, and give the clip an explicit arc across its full length.

Frequently asked questions

How long can a Seedance 2.5 clip be?

5 to 30 seconds, selectable in whole seconds, with 5 as the default. Five seconds is also the minimum billed length, so a shorter request is not possible here.

Does Seedance 2.5 have 1080p?

No. The available tiers are 480p and 720p, with 480p as the default. Seedance 2.0 Pro is the tier in this family that offers 1080p, in exchange for a 15-second ceiling on duration.

What is the adaptive aspect ratio?

A seventh option alongside 16:9, 4:3, 1:1, 3:4, 9:16 and 21:9, in which the output follows the framing of the reference material instead of a fixed frame. Pick a fixed ratio when the delivery has a required frame size.

What reference material does it accept?

Up to 9 reference images, up to 3 reference videos totalling 30 seconds, and up to 3 reference audio tracks alongside a reference image or video. The 30-second video budget is twice what the Seedance 2.0 tiers allow.

How is Seedance 2.5 billed on Genvizu?

Per second of output, with a five-second minimum. A run that includes a reference video uses a lower per-second rate but counts the input seconds as well as the output seconds. Current tiers are on the pricing page, and the generate button shows the exact cost before you confirm.

Does it generate sound?

Yes, audio is generated by default and can be switched off. Describing the sound you want in the prompt gives a far better result than leaving it unspecified.

Sources