What is Seedance?

Seedance in one paragraph

Seedance is not an app you install. It is a set of generative video models that other products call through an API. ByteDance Seed publishes each version with an official page and a launch post; access then comes through ByteDance's own channels and through third-party platforms that license the models. Seedancee is one of those independent platforms: we run Seedance models through the Kie.ai provider inside a creation workspace, and we are not affiliated with ByteDance.

The model family, version by version

As of 2026-09-29, the Seedance versions you are most likely to encounter are Seedance 2.5 (the current flagship), Seedance 2.0 and its Fast and Mini variants, Seedance 1.5 Pro, and the original Seedance 1.0. The table lists what is available in the Seedancee workspace; capability numbers come straight from our model registry, which we check against the provider's documentation.

Seedance 2.5

ByteDance launched Seedance 2.5 on 2026-07-31 and describes it as "30-second storytelling with precise reference control and powerful editing". In practice that means three things: clips can run up to 30 seconds in a single generation, the model accepts a large set of reference images, video and audio (up to 30 images, 10 videos, 10 audio clips on Seedancee), and finished clips can be edited and extended rather than regenerated from scratch. The official page also describes "white-model" control, where a plain 3D block-out guides layout and motion, and green-screen editing for replacing backgrounds.

Seedance 2.0, 2.0 Fast and 2.0 Mini

Seedance 2.0 launched on 2026-02-12. It produces 4–15s clips at up to 4k and introduced multimodal referencing: images, short video clips and audio can all steer the result. Seedance 2.0 Fast trades the highest resolutions for quicker turnaround, and Seedance 2.0 Mini is the lowest-cost option for drafts and social clips. All three share the same reference limits (up to 9 images, 3 videos, 3 audio clips).

Seedance 1.5 Pro and 1.0

Seedance 1.5 Pro is the previous generation still offered by the provider. It handles text-to-video and image-to-video with first and last frame control, generates native audio, and has a "fixed lens" option that keeps the camera still. It does not accept reference images, video or audio. Seedance 1.0 is the original release; we do not offer it in the workspace and we do not list a launch date for it because we could not verify one from a primary source.

How Seedance generates a video

Every Seedance workflow starts with a prompt and, optionally, some reference material. The model turns that into a clip with a fixed duration, resolution and aspect ratio. The three core modes are:

1. 01

   Text to video

   You describe the subject, action, environment, camera and lighting in words. The model invents everything. Best for concept shots, mood pieces and B-roll where you do not need a specific face or product.

2. 02

   Image to video

   You supply a first frame (and optionally a last frame). The model animates from that frame, preserving composition and colour. Best for product shots, illustrations you want to bring to life, and consistent openings.

3. 03

   Reference to video

   You supply reference images, video or audio and tell the model how to use them: keep this character, copy this camera move, match this voice. Seedance 2.0 and 2.5 support this; 1.5 Pro does not.

Seedance 2.5 adds two more modes on top of a finished clip: edit, where you change part of a generated video, and extend, where the model continues a clip. ByteDance's page says a clip can be extended up to twice.

What native audio means

Older video models produced silent clips and left sound to a separate step. Seedance generates audio in the same pass: ambient sound, effects tied to on-screen action, and speech with lip movement when the prompt contains dialogue. Reference audio on 2.0 and 2.5 lets you supply a voice or a beat for the model to match. Our native audio feature page covers how to prompt for it and where it still falls short.

Where you can use Seedance

ByteDance's official Seedance 2.5 page names three access channels: the Dola chat interface, the BytePlus Playground, and an API. Third-party platforms such as Kie.ai license the models and expose them through their own APIs; Seedancee builds on Kie. The model weights are not publicly released as far as we can verify, so there is nothing to download and run locally. Our access guide keeps a dated list of channels.

What Seedance is not

• Not an open-source project: no public weights or training code as of the verified date.
• Not a desktop or mobile app: any site offering a 'Seedance download' is offering something else.
• Not a character or brand cloning tool: reproducing copyrighted characters or real people without consent violates our content policy and, in many cases, the provider's moderation.
• Not a guarantee of a specific output: generation is probabilistic; the same prompt produces different clips each time.