All posts

SEEDANCE · AUGUST 7, 2026 · UPDATED AUGUST 28, 2026 · 6 MIN READ

Seedance 2.5 Is Here: 30-Second Clips, 30 References, and What Changed From 2.0.

Seedance 2.5 is ByteDance’s live upgrade to Seedance 2.0 — 30-second clips, up to 30 references, and a new Edit/Extend mode. What changed, and how to use it on getvivix.

getvivix Team
getvivix Journal
August 7, 20266 min

Seedance 2.5 is live — ByteDance’s next step up from Seedance 2.0, and it’s available now in getvivix Studio. Where Seedance 2.0 Mini is the budget tier of the family, 2.5 is the scale tier: far more reference material per generation and clips more than twice as long, on the same multimodal engine that made 2.0 reliable in the first place.

It’s a genuinely bigger jump than a version-number bump suggests. The two things that used to give AI video away — a hard length ceiling and characters drifting over a long take — are exactly what 2.5 was built to push past.

Key Takeaways

  • Seedance 2.5 renders single-take clips up to 30 seconds — double Seedance 2.0’s 15-second ceiling.
  • Reference capacity jumps to up to 30 images, 10 videos, and 10 audios in one generation (2.0: 9 / 3 / 3).
  • A new Edit/Extend mode reworks an existing clip in place or extends it past its last frame — 2.0 doesn’t have this.
  • Resolution moves the other way: 2.5 renders natively up to 1080p, but Seedance 2.0 can still go higher, to native 4K.
  • Live now in getvivix Studio alongside the rest of the Seedance family, credit cost shown before every generation.

What actually changed from Seedance 2.0

Two numbers matter more than any others here: how long a single take can run, and how much reference material you can feed the model in one generation. Here’s exactly what moved, per ByteDance’s own BytePlus documentation. Source: ByteDance BytePlus.

  • 30-second single-take clips — double the 15-second ceiling on Seedance 2.0, in one continuous render, not a stitched sequence.
  • Up to 30 reference images — versus up to 9 on 2.0. Enough for a full cast, wardrobe, and location set in a single generation.
  • Up to 10 reference videos — versus up to 3 on 2.0, for motion and continuity references.
  • Up to 10 reference audios — versus up to 3 on 2.0, for voice and sound-driven generation.
  • A rebuilt Omni mode — text, image references, video references, and audio references combine freely in one request instead of separate modes for each.
  • A dedicated Frame to Video mode — set a first frame and an optional last frame when you need precise start-and-end control, kept separate from Omni’s free-form referencing.
  • A new Edit/Extend mode — upload an existing clip and a prompt, and the same request either edits it in place (swap or remove a subject, same length) or extends it past its last frame, up to a 30-second total. 2.0 doesn’t have this mode at all.

What didn’t improve in 2.5’s favor: resolution. Seedance 2.5 renders natively up to 1080p — up from a 720p ceiling at its own launch — but Seedance 2.0 can still go higher, rendering native 4K. If your read on “2.5” was a straight resolution upgrade, it isn’t; 2.5’s jump is capacity, duration, and editing, not maximum resolution. Native multimodal input (text, image, reference, audio) carries over from 2.0 unchanged.

Seedance 2.5 vs Seedance 2.0, side by side

SpecSeedance 2.0Seedance 2.5
Max clip length15 seconds30 seconds
Reference imagesUp to 9Up to 30
Reference videosUp to 3Up to 10
Reference audiosUp to 3Up to 10
Max resolutionUp to 4K (native)Up to 1080p (native)
Generation modesText / image / reference / audio-to-video (separate tabs)Omni (all references combined) + Frame to Video + Edit/Extend

Where 2.0 still wins: if you just need a quick, cheap iteration on a short clip, Seedance 2.0 Mini and Seedance 2.0 Fast are lighter and faster to render. And if native 4K is the actual requirement, that’s Seedance 2.0’s ceiling, not 2.5’s — 2.5 is built for the shot that needs to hold a character, a set, and a full 30 seconds together in one take, up to its own 1080p cap — that’s where the extra reference headroom actually earns its keep.

How to use Seedance 2.5 on getvivix

Seedance 2.5 is in the Studio model list now, alongside every other model in the Seedance family. Three ways to work with it:

  • Omni — drop in any mix of reference images, videos, and audio (all optional — a plain text prompt works too), write your prompt, and generate. This is the mode for anything that needs consistent characters, wardrobe, or setting across the full clip.
  • Frame to Video — upload a first frame (and optionally a last frame) when you need the shot to start and end on exact compositions.
  • Edit/Extend — upload an existing clip instead of starting from scratch, and write a prompt describing the edit or the extension you want.

getvivix shows the credit cost before you hit generate, on every model, every time — no surprise at checkout. Seedance 2.5 sits in the same affordable range as the rest of the family, and there’s current launch pricing on it — worth a look at the pricing page before your first generation.

Where the engine comes from

Seedance runs on ByteDance’s multimodal video engine, documented directly for API access on ByteDance’s own BytePlus platform. getvivix runs it through a visual Studio composer instead: reference uploaders, a prompt box, and one credit balance shared across every model, so you never touch a raw API call to use it.

Try Seedance 2.5 on getvivix free — trial credits on signup, no card required.

NEXT IN JOURNAL

Sora 2 Prompts: 30 Examples That Actually Work (2026)
Read

RELATED READING

Newsletter

Be the first to know

Subscribe to the getvivix newsletter and you'll hear it first whenever new models land or new features go live. No promo spam. Unsubscribe in one click.

We use your email only for the newsletter. Unsubscribe anytime.