Seedance 2.5 is an audio-video generation model developed by ByteDance that produces clips up to 30 seconds long. Developers and video editors use it to create consistent scenes by providing text prompts alongside image, video, or audio references. It solves the common problem of short, 5-10 second generation limits found in many other AI video tools.
The model handles complex instructions for camera movement and lighting while maintaining subject consistency across longer durations. It works through the Wan API, allowing for integration into custom workflows or testing within a playground environment. One specific advantage is the ability to generate sound effects or dialogue that match the visual action in a single pass.
Key Features
- Maximum duration of 30 seconds for a single video generation
- Synchronized audio generation that matches the visual content
- Support for up to 30 image references to define character or location appearance
- Reference video input for 10 clips to guide specific motion or timing
- Resolution options including 480p, 720p, and 1080p outputs
- Adaptive aspect ratios that follow the prompt or input media
- Draft mode for 480p previews before committing to high-resolution final renders
- Last-frame image return to assist with chaining multiple shots together
How to Use
- Access the Wan API playground and select the Seedance 2.5 model.
- Write a detailed shot brief describing the subject, action, and camera movement.
- Upload image or video references to guide the visual style and character consistency.
- Toggle the audio generation setting if you need environmental sounds or music.
- Choose your output resolution and aspect ratio before clicking generate.
- Review the 480p draft and then proceed to the 1080p final version if the motion is correct.
Use Cases
- Creating short narrative scenes with a clear beginning, middle, and end.
- Developing product motion studies to test how light interacts with specific materials.
- Building previsualization clips for film projects to explore framing and timing.
- Generating social media content that requires synchronized background audio and visuals.
Pricing
Pricing is calculated per billable second and varies by resolution. For 1080p video without a reference, the rate is approximately $0.5971 per second, while 480p starts around $0.108 per second. Reference videos add to the billable input time.
FAQ
What is Seedance 2.5?
Seedance 2.5 is a ByteDance model designed for generating longer video clips with integrated audio and high reference capacity.
Is Seedance 2.5 free?
No, it uses a pay-per-second pricing model through the Wan API.
Can I generate 4K video?
No, the current maximum native resolution for Seedance 2.5 is 1080p.
How many images can I use as references?
You can submit up to 30 images to guide the model on appearance and style.
Does it support audio-only input?
You can provide up to 10 audio clips as references to guide the generated sound.
What is the maximum video length?
A single generation can last up to 30 seconds.
