AI Video and Image Models on SeaDance

Choose from seven video and four image models. These cards describe the options offered on SeaDance; model news below covers models that are not connected to the generator.

Video models

Seedance 2.5

One 16:9 video per request at 480p, lasting 5, 10 or 15 seconds.

Supported modes: Text to Video · Image to Video · Reference to Video

Reference image limit: Image to Video: 1–2; Reference to Video: 1–9

Seedance 2.0

One 16:9 video per request at 480p, lasting 5, 10 or 15 seconds.

Supported modes: Text to Video · Image to Video · Reference to Video

Reference image limit: Image to Video: 1–2; Reference to Video: 1–9

Seedance 2.0 Fast

One 16:9 video per request at 480p, lasting 5, 10 or 15 seconds.

Supported modes: Text to Video · Image to Video · Reference to Video

Reference image limit: Image to Video: 1–2; Reference to Video: 1–9

Seedance 2.0 Mini

One 16:9 video per request at 480p, lasting 5, 10 or 15 seconds.

Supported modes: Text to Video · Image to Video · Reference to Video

Reference image limit: Image to Video: 1–2; Reference to Video: 1–9

Kling 3.0

One 16:9 video per request at 720p, lasting 5, 10 or 15 seconds.

Supported modes: Text to Video · Image to Video

Reference image limit: Image to Video: 1

MiniMax H3

One 16:9 video per request at 768P, lasting 5, 10 or 15 seconds.

Supported modes: Text to Video · Image to Video · Reference to Video

Reference image limit: Image to Video: 1–2; Reference to Video: 1–9

MiniMax H3 Max

One 16:9 video per request at 768P, lasting 5, 10 or 15 seconds.

Supported modes: Text to Video · Image to Video · Reference to Video

Reference image limit: Image to Video: 1–2; Reference to Video: 1–9

Image models

GPT Image 2.5

One 1K / 2K / 4K image per request; image editing accepts up to 16 reference images.

Supported modes: Text to Image · Image to Image

Reference image limit: 16

GPT Image 2

One 1K image per request; image editing accepts up to 9 reference images.

Supported modes: Text to Image · Image to Image

Reference image limit: 9

Nano Banana 2.1

One 1K / 2K / 4K image per request; image editing accepts up to 10 reference images.

Supported modes: Text to Image · Image to Image

Reference image limit: 10

Nano Banana Pro

One 1K / 2K / 4K image per request; image editing accepts up to 4 reference images.

Supported modes: Text to Image · Image to Image

Reference image limit: 4

Model news and status

These are information pages, not generation options. Their release or preview status does not mean you can use them on SeaDance.

Model questions

Which models can I use on SeaDance?

The seven video and four image models listed above are available for generation. Choose a model on its page or in the video or image generator.

Does every video model support reference images?

All seven support Image to Video. Reference to Video supports up to nine images on the Seedance and MiniMax models; Kling 3.0 supports a single start image and has no Reference to Video mode.

How many reference images can I use for image editing?

GPT Image 2.5 accepts up to 16, GPT Image 2 up to 9, Nano Banana 2.1 up to 10 and Nano Banana Pro up to 4 reference images in Image to Image.

Can I generate with the models in the news section?

No. Those six pages describe model news and status. The models are not available on SeaDance and cannot be selected for generation here.