ByteDance's Seed team has officially announced its video generation model "Seedance 2.5" and simultaneously updated the existing "Seedance 2.0" to support 4K output. It also revealed a copyright commercialization platform built on officially licensed IP.
Jun 23, 2026 · ByteDance / Volcano Engine
Seedance 2.5 Pushes AI Video to 30-Second, 4K, Single-Prompt Clips
ByteDance unveiled its next-gen video model at the FORCE Conference in Beijing — supporting up to 50 multimodal reference assets, 3D white-model input, and native 4K output, with general availability planned for early July 2026.
30s
Single-segment clip from one prompt
4K
Native output resolution
50
Multimodal reference assets supported
$0.15
Generation cost per second (API)
2.0 vs 2.5 — drawn to scale
Each column built from stacked blocks; taller column = larger value.
MAX CLIP LENGTH (seconds)
What's new in 2.5
3D WHITE MODELS
Accepts white models exported from tools like Blender as reference input.
COPYRIGHT PLATFORM
New AI copyright commercialization platform with licensed IP and revenue sharing.
MULTIMODAL INPUT
Up to 50 images, video, audio and text references in a single generation.
Strengths cited
High-quality footage from one image + prompt
Professional-grade shot transitions
Strong character consistency & lip-sync
Good quality-to-price balance
Open constraints (current 2.0)
~15-second length limit
Lip-sync drift on long dialogue
Limits with photorealistic human references
Cost noted as high for heavy use
Timeline:
Seedance 1.0 (Jun 2025) → Seedance 2.0 official (~Feb 2026) → Seedance 2.5 beta now, general release planned early July 2026.
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…