Breaking

ByteDance's Seedance 2.5 Heads to CapCut With Up to 30-Second Video

July 4, 2026 at 05:14 EDT

Media Generation
Foundation Models

ByteDance's AI video-generation model "Dreamina Seedance 2.5 (Seedance 2.5)" is set to arrive soon on the video-editing app CapCut and partner apps including Dreamina. In standard mode it can generate cinematic clips of up to 30 seconds, while a beta long-video mode extends clips further.

Early–Mid July 2026 · ByteDance

Dreamina Seedance 2.5 Brings Longer, Sharper AI Video to CapCut

ByteDance's next-gen video model generates cinematic clips up to 30 seconds — stretching to 3 minutes in beta long-video mode — with up to 50 multimodal reference inputs and region-level editing built into CapCut.

30s

standard clip length

180s

long-video mode (beta)

reference inputs (text·image·video·audio)

output support

Reference inputs: 2.0 → 2.5

Roughly 4× more reference inputs enable more complex, controllable storytelling.

~12

Seedance 2.0

Seedance 2.5

What's new in 2.5

Item

Seedance 2.0

Seedance 2.5

Clip length

Shorter

Up to 30 seconds

Long-video mode

—

Up to 180s (beta)

Reference inputs

Around 12

Up to 50

Editing

Basic generation

Region editing, R2V, better character consistency

Strengths (carried from 2.0)

Top-tier prompt adherence & natural output
Strong character consistency for solo creators
Seamless generate-and-edit inside CapCut

Open concerns

Limits on real-person face generation
Slow, costly generation & unwanted added music
Questions over subscription value & support

Staged Rollout

China first

priority markets

→

CapCut & Dreamina

Web · Desktop · Mobile

→

Mini & Fast

planned versions

Positioned against Kling, Sora and Veo — differentiated by an integrated generate-and-edit workflow and multimodal control for short dramas, ads and social content.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.