Seedance 2.5: Video Settings, Pricing and Sources

Seedance 2.5 is ByteDance's video generation model. On Effect Video you enter a text prompt, optionally add one reference photo, and choose a length of 4, 5, 8, 10, 15 or 30 seconds at 480p, 720p or 1080p, with the total credits shown before you generate. This guide also covers what is publicly documented about the model and how creators have tested it.

What Seedance 2.5 does

ByteDance Seed introduced the model in a post titled One-Take Creation, Flexible Referencing, which points to one-take generation and flexible referencing as its themes. The KIE Seedance 2.5 interface documentation describes the video generation interface that Effect Video builds on.

On the BytePlus channel, the Dreamina Seedance 2.5 overview video (July 31, 2026) is titled "Built for Professional AI Video Production". On August 14, 2026, Dreamina AI announced that Dreamina Seedance 2.5 is available in 1080p, making its global debut on Dreamina Web across Southeast Asia, the Middle East, Africa, Europe and South America. On September 23, 2026, Dreamina AI announced the winners of its White Model Reference Creation Contest, where creators started from the same white-model assets and used the model's White Model Reference capability.

What creators and the community say

On YouTube, Curious Refuge (July 31, 2026) shared an honest opinion on Seedance 2.5 under the title "Seedance 2.5 Is Incredible... Until It Isn't". ElevenLabs (August 2, 2026) published an overview of Seedance 2.5 versus 2.0, and Dan Kieft (August 7, 2026) posted a full tutorial.

On X, Alpha Mom (October 9, 2026) ran Seedance 2.5 and Wan 3.0 on the same fitness vlog prompt, and wrote that "Seedance looks much more natural to me." Their X profile lists makeugc as a partner, and the test was run on that platform. Rohit (October 9, 2026) shared results from World Models' Last Exam in Physics, reporting 57.76 out of 100 for Seedance 2.5 across 40 tasks. We have not verified this score, and the post does not give the benchmark method or data, so it is not a basis for choosing a model.

On Reddit, threads touch on prompting, realism, consistency and cost; the other posts and threads are shown further down this page and in the references.

Using it on Effect Video

Seedance 2.5 is available on Effect Video: open the generator with Seedance 2.5 selected. You provide a text prompt, optionally add one reference photo, and choose a length of 4, 5, 8, 10, 15 or 30 seconds at 480p, 720p or 1080p. Sound can be switched on or off; it does not change the credit price. The generator shows the total credits before you start, and finished videos stay in your account for 7 days, so download what you want to keep.

Limits and open questions

• Reference inputs: Effect Video accepts a text prompt and one reference photo. Upstream services may offer additional reference inputs, which this generator does not expose.
• Evaluation contexts: The benchmark scores and side-by-side tests above use different prompts and methods, so treat them as individual data points rather than a general ranking.
• Sources: The ByteDance Seed announcement is rendered client-side, so this guide relies on its title and link rather than its full text. No independent written review has been verified yet.

For consistent results, pair a clear descriptive prompt with a clear reference photo.