Seedance 2.5 is ByteDance's newest AI video model, and every video feature on Photo AI now runs on it. This guide explains what the model can do, what changed compared to Seedance 2.0, and how to get the best videos of your AI character out of it.

What is Seedance 2.5?

Seedance 2.5 turns a text prompt, an image, or a set of reference photos into a cinematic video clip with sound. Two things set it apart from most video models: it generates up to 30 seconds in ONE continuous take, so you get a coherent shot instead of stitched segments, and the audio (ambience, footsteps, dialogue, music) is generated together with the picture in the same pass, already in sync.

What's new compared to Seedance 2.0?

Clips are twice as long (up to 30 seconds in one take), prompt adherence is noticeably better for camera moves and timing, faces and lighting stay consistent through the whole clip, and it accepts far more reference images, clips and audio tracks, which it actually uses to lock identity, products, locations and style.

What runs on Seedance 2.5 in Photo AI?

Everything video: [ Make a video ] in the sidebar (a video starring your character straight from a prompt, no photo needed first), [ Make video ] on any photo in your gallery (image-to-video), [ Extend video ], video [ Magic edit ] (describe a change and the whole video re-renders with it), and the Video Editor where you can generate clips on a timeline and let the AI agent assemble your story.

How do I make a video of my character?

Select your character on photoai.com (or create one first, creating characters is free), open the [ Make a video ] tab, and write a prompt describing what happens. Write model to reference your character, for example: "model walking through Tokyo at night in the rain, neon signs reflecting in puddles, camera slowly orbiting". Pick a length (5, 10, 15 or 30 seconds) and press [ Make video ]. The video appears in your gallery a few minutes later, with sound. You can select multiple characters to star together, and batch generate by separating prompts with |||.

Can I add my own reference photo or voice?

Yes. Below the prompt box there are two optional uploads: a reference photo (any photo, your living room, a beach, an outfit, a style frame, used as the scene or style of the video) and a voice sample (mp3, mp4, aac, wav, m4a or ogg, your character will speak with that voice).

How do I write good video prompts?

Write it like a shot list: subject, action, setting, camera movement, lighting, mood. "model pouring coffee in a sunlit kitchen, slow push-in, warm morning light, relaxed" beats a pile of adjectives. Let the references define the look, your character's photos already carry the face and hair, so spend your words on action, wardrobe and camera. Direct the audio too (rain, a busy street, the character speaking). And keep one idea per clip: several focused shots cut together in the Video Editor beat one clip trying to do five things.

What does it cost?

Video pricing scales with length: a 15 second video costs 30 credits, so 5s is 10 credits, 10s is 20, and a full 30 second single-take clip is 60 credits. That makes Photo AI the cheapest way to use Seedance 2.5 right now, generating the same clip through video APIs directly costs several dollars, and most video platforms charge multiples of our credit price.

Do my videos have a watermark?

No, videos download clean.

Can I use them commercially?

Yes, videos you generate with your own characters are yours to use, including for commercial content.

How long does a video take?

Usually a few minutes. The [ Make video ] button shows the live average for your selected length.

Which aspect ratios are supported?

Your Orientation preference is used: portrait, 9:16 for TikTok, square, landscape, 16:9 for YouTube, and cinematic widescreen.

Do I need to generate a photo first?

Not anymore. [ Make a video ] goes straight from your prompt plus your character's reference photos to a finished video.