Seedance 2.5 Launched: Video Length Doubled to 30 Seconds, Multimodal References Fully Upgraded

Seedance 2.5 Launch: Video Length Doubled to 30 Seconds, Multimodal References Fully Upgraded

ByteDance's new-generation video model Seedance 2.5 is now live on Genvizu. The most noticeable change this time is the doubling of the maximum generation length from 15 seconds to 30 seconds. Simultaneously, all three reference channels—image, reference video, and reference audio—are now fully integrated, and a new adaptive aspect ratio option has been added. Here's a quick guide on how to use the new features and understand the pricing.

What Fits in 30 Seconds

• Generate videos from 5–30 seconds (previously Seedance 2.0 had a 15-second limit). This allows for complete short narratives with setup, climax, and resolution, moving beyond single actions or transitions.

• Fixed resolutions: 480p / 720p (no 1080p / 4K). Computational power is prioritized for length rather than ultra-high definition.

• 7 Aspect Ratio Options: 16:9 / 4:3 / 1:1 / 3:4 / 9:16 / 21:9, plus the new adaptive (default) option—when unsure whether the final video should be landscape or portrait, the model intelligently determines the composition based on content.

• Start and End Frame Control is retained, allowing precise locking of opening and closing scenes even in 30-second videos.

Multimodal References: Image + Reference Video + Reference Audio, Fed to the Model Together

Seedance 2.5 expands reference inputs to three channels. Reference images can even be specifically selected within the prompt using their numbers:

• Up to 9 reference images

• Up to 3 reference videos, with a total length ≤30 seconds

• Up to 3 reference audio clips, with a total length ≤30 seconds

In the prompt, you can directly use numbers like @Image1 to specify the role of a reference image, for example:

Reference @Image1 @Image2 for the main character, @Image3 for the scene.
Keep the outfit from @Image1 throughout the video.

No more lengthy text descriptions like "first image" or "second image." The model directly associates the number with the corresponding image. Reference videos and audio do not require numbering; once uploaded, they automatically serve as references for motion rhythm and sound texture. Audio generation is also an independent toggle—turn it off for scenes that don't require voiceovers or music, outputting silent videos.

How Pricing Works: Two Tracks, Different Prices and Calculation Methods

Seedance 2.5 uses two pricing tracks based on resolution and whether reference video is used:

TierWithout Reference VideoWith Reference Video (Calculated by Total Input + Output Seconds)480p35 Credits/Second22 Credits/Second720p79 Credits/Second48 Credits/Second

These two track prices cannot be directly compared because the calculation of seconds differs: Without reference video, billing is based solely on output seconds; with reference video, the price per second is lower, but billing is based on the total seconds of "reference video input duration + output duration." The minimum billing is 5 seconds.

Example: 480p, 5 seconds of text-to-video (without reference video) = 35 × 5 = 175 Credits; 720p, 30-second video = 79 × 30 = 2370 Credits; 480p, 5 seconds output + one 10-second reference video = 22 × (10 + 5) = 330 Credits. For scenarios with reference video, carefully check if billing is based on total seconds before deciding whether to shorten the reference video.

Seedance 2.0 vs. Seedance 2.5: Which to Choose

These two versions are not replacements; choose based on your scenario:

• Seedance 2.0 (Fast / Pro / Mini): Maximum length 15 seconds, total reference video length ≤15 seconds. Fast / Mini are quick and low-cost, suitable for short clips, transitions, and rapid iteration; Pro offers 1080p high-definition.

• Seedance 2.5: Length doubled to 30 seconds, reference video length doubled to 30 seconds (reference audio length also 30 seconds), new adaptive aspect ratio. Ideal for projects requiring complete narrative pacing or simultaneous input of multiple images, videos, and audio assets.

To refine your visual language, the layered approach to subjects, actions, shots, and sounds described in Seedance 2.0 Text-to-Video Prompt Tutorial and Seedance 2.0 Image-to-Video Prompt Tutorial is still applicable to 2.5, now with more duration and reference material to leverage. For earlier updates on reference video/audio capabilities and pricing changes, refer to Seedance 2.0 Full Upgrade: Reference Video and Audio Support.

Try It Now

Whether you want to tell a complete 30-second story or precisely direct multiple reference images using @Image1, complemented by reference video and audio assets, these capabilities are now available directly on Genvizu: Open the Genvizu AI Infinite Canvas