video model by ByteDance
Seedance 2.5
Up to 30 seconds of video from a prompt, frames or references
What is Seedance 2.5?
Seedance 2.5 is a video model from ByteDance. In Cinemagiq it comes in three versions: Seedance 2.5 for prompts and frames, Seedance 2.5 References for shots built from reference material, and Seedance 2.5 Live Action, which keeps real-looking AI characters from up to 9 reference images. It is the only model in Cinemagiq that reaches 30 seconds in a single generation, which makes it a natural pick for longer, story-driven moments.
A prompt alone gives you text-to-video. Add a first frame, a last frame or both to pin where the shot starts and ends. Seedance 2.5 References takes up to 30 images, 10 videos and 10 audio clips, but frames and references cannot be combined in one job. Clips run 4 to 30 seconds at 480p, 720p or 1080p, in ratios from 21:9 to 9:16 or adaptive, with sound you can switch on or off.
Pick Seedance 2.5 when a beat needs room to breathe: a walk-and-talk, a chase, or several shots cut together in one prompt. ByteDance documents multi-shot prompts, and Cinemagiq automatically rewrites your Seedance prompt into the structure ByteDance recommends. Every clip lands in the Asset Hub under All Generations, so you can stack takes on a shot in the Production Board and star the one you keep. For clips of 15 seconds or less, compare it with Seedance 2.0 and Kling O3.
Seedance 2.5 examples
Made in Cinemagiq with Seedance 2.5. The exact prompt is under each one.
PromptShot 1: wide shot of a rain-soaked neon alley at night, steam rising from a grate, a woman in a yellow raincoat walks away from the camera. Shot 2: close-up of her face lit by a flickering pink sign as she looks back over her shoulder. Sound: rain, distant traffic, the buzz of the neon sign, no music.
PromptA slow dolly-in on an old wooden sailboat in a misty harbor at dawn, gulls taking off from the mast, water gently lapping against the hull. Sound: water, creaking wood and gulls, no music.
What Seedance 2.5 can do in Cinemagiq
Straight from the model list Cinemagiq runs today.
| Variant | What drives the shot | Length | Resolution | Sound |
|---|---|---|---|---|
| Seedance 2.5Prompt, frames or references drive the shot, up to 30s with audio | Prompt, first frame (optional), last frame | 4 to 30 s | 480p, 720p, 1080p | Optional |
| Seedance 2.5 ReferencesPrompt, frames or references drive the shot, up to 30s with audio | Up to 30 reference images, plus reference videos and audio | 4 to 30 s | 480p, 720p, 1080p | Optional |
| Seedance 2.5 Live ActionReal-like AI characters from reference images, up to 30s with audio | Up to 9 reference images, plus reference videos and audio | 4 to 30 s | 480p, 720p, 1080p | Optional |
Strengths
- Up to 30 seconds in one generation
- Multi-shot prompts with Shot headings or a timeline
- Up to 30 reference images, 10 videos and 10 audio clips
- Optional sound with the picture
- Live Action keeps real-looking characters from references
Best for
- Long takes and scene-length beats
- Multi-shot sequences from one prompt
- Character scenes built from reference images
- Shots that need sound in the same pass
Watch-outs
- Frames and references cannot be combined in one job
- Seedance 2.5 Live Action works from reference images only, with no first or last frame
- The shortest clip is 4 seconds
Prompt tips for Seedance 2.5
- For several shots, use Shot 1, Shot 2 headings or a gapless timeline in whole seconds (0-3s, 3-8s). For one continuous take, write a single block with neither.
- Add a Sound line that lists what we hear, like the alley example on this page: rain, distant traffic, the buzz of the neon sign, no music.
- With a first frame, spend the prompt on motion and camera, not on what is already in the picture.
- In Seedance 2.5 References, say what each reference is for (a character, a location, a prop) so it guides the right part of the shot.
How to use Seedance 2.5 in Cinemagiq
- 1Open the Generation Studio in your project and switch to the Video tab.
- 2Pick Seedance 2.5 in the AI Model picker. Each card says what drives the result and its limits.
- 3Write the prompt, and add a First Frame, Last Frame or reference images where the model takes them.
- 4Generate. The button shows the token cost before you start.
- 5Find every take in the Asset Hub under All Generations, and add the one you keep to its shot on the Production Board.