Text to video
Seedance 2.5
ByteDance’s video model for 30-second 4K clips with audio, guided by up to 50 references.
Product insights
Seedance 2.5 at a glance
Value proposition
Get a 30-second clip with audio in one generation instead of stitching several short shots together.
Problem solved
Earlier video models capped clips at a few seconds, so longer scenes needed stitching and drifted between shots. Seedance 2.5 keeps one continuous take for longer and lets you fix parts of it without regenerating the whole clip.
Audience
Filmmakers, marketing and ecommerce teams, social media creators, and agencies.
Pricing and access
Dreamina gives free daily credits to try it. Paid Dreamina plans use credits, with launch pricing from about $0.097 per second of Seedance 2.5 video. Prices vary by region and plan.
Market
- AI video generation
- Short-form ads and social content
- Previs and filmmaking
Tech stack
- ByteDance Seed video model
- Joint audio and video generation
- Dreamina · CapCut · Jimeng · Doubao
Model sheet
The working parts of Seedance 2.5
- Maker
- ByteDance Seed
- Released
- July 31, 2026
- Clip length
- Up to 30 s per run; 3 min in long-video mode (beta)
- References
- Up to 50: 30 images, 10 videos, 10 audio
- Where to use
- Dreamina, CapCut, Jimeng, Doubao
01Why it is here
What is Seedance 2.5?
Seedance 2.5 is the July 2026 update to ByteDance’s Seedance video models. Compared with Seedance 2.0, it raises the single-run length to 30 seconds, adds a 3-minute long-video mode in beta, and accepts up to 50 multimodal references, including images, video clips, and audio. Sound is generated together with the picture. An Intelligent Edit mode lets you change part of a clip, and green-screen and 3D white-model inputs (including Blender and Maya assets) help with precise motion and compositing. Dreamina and CapCut offer it worldwide; Jimeng and Doubao carry it in China. Use reference photos only of people who have agreed to it.
02Workflow
How to use Seedance 2.5
- 01
Open Seedance 2.5
Sign in to Dreamina (or Jimeng in China) and choose Seedance 2.5 as the video model.
- 02
Add a prompt and references
Describe the scene, then add reference images, video, or audio for characters, motion, and style.
- 03
Generate and edit
Generate up to 30 seconds, then use edit mode to fix parts or extend the clip.
03Capabilities
Key features of Seedance 2.5
- 30-second single-run clips
- Up to 50 multimodal references
- Audio generated with the video
- Intelligent Edit, green screen, and 3D white-model control
04Use cases
What is Seedance 2.5 best for?
- Product ads and ecommerce videos
- Short film scenes and previs
- Social clips with consistent characters
05Questions
FAQ about Seedance 2.5
What is new in Seedance 2.5?
Longer clips (30 seconds per run, 3 minutes in beta long-video mode), up to 50 references instead of 12, partial editing, and green-screen and 3D white-model control.
Where can I use Seedance 2.5?
On Dreamina and CapCut worldwide, and on Jimeng and Doubao in China. Third-party studios such as Ausar also offer it.
Is Seedance 2.5 free?
Dreamina gives free daily credits to try it. Regular use needs a credit plan; launch pricing starts at about $0.097 per second of video.
Official website snapshot
dreamina.capcut.com
Seedance 2.5 is ByteDance Seed’s video generation model, released on July 31, 2026. It generates up to 30 seconds of continuous video in one run, with a beta long-video mode that extends to 3 minutes, and produces sound together with the picture. You can guide it with up to 50 references (30 images, 10 videos, and 10 audio clips), edit a specific part of a result, or use green-screen and 3D white-model inputs for tighter control. It is available on Dreamina and CapCut worldwide and on Jimeng and Doubao in China.
