Clips up to 30 seconds
Room for a full beat: setup, action, and payoff.
Hot
ByteDance’s Seedance 2.5 engine — strong motion, multi-reference conditioning, and synchronized audio on clips up to 30 seconds.
Avoid copyrighted characters, logos, music and sensitive or explicit content — it may be blocked.
Estimated cost: 5 credits
Your creation appears here
Pick a model, describe the shot, and hit Generate.
Why Seedance 2.5
Seedance 2.5 is the Seedance model for long clips, first frames, and big reference sets.
Room for a full beat: setup, action, and payoff.
Hold a character and style steady across a sequence.
Switch on sound that is generated with the picture.
Start from a still to lock the look before motion.
Seedance 2.5 guide
Seedance 2.5 is ByteDance’s video generation model, available in the Nano Banana AI studio next to Muse Video, Veo 3.1, Wan 3.0, Kling 3.0, and MiniMax H3. It is the pick when you need longer clips, many reference images, and synchronized sound in one generation. This guide explains what each input does and how to get a usable clip on the first few tries.
Seedance 2.5 handles three jobs: text to video, image to video from a first frame, and reference to video. In text mode you describe the shot. In image mode you supply the opening frame and describe how it moves. In reference mode you supply up to 30 reference images, plus up to 10 reference videos and 10 audio clips, so the model keeps a character, outfit, or visual style steady across the clip. A last frame can be added to a first frame to define where the shot ends.
Clips run from 4 to 30 seconds at 480p, 720p, or 1080p, which makes the model unusually good for scenes that need time to develop rather than a single looping gesture. An Audio toggle adds synchronized sound generated together with the picture, so you are not left matching a soundtrack afterwards.
Durations from 4 to 30 seconds let a single generation carry an establishing shot, an action, and a payoff.
Up to 30 reference images hold identity and style, which is the difference between one lucky clip and a repeatable character.
Choose 16:9, 9:16, 1:1, 4:3, 3:4, 21:9, or adaptive to match the source image.
Think in shots, not in sentences. A good video prompt states the subject, the action, the camera, and the light: "A courier cycles through a rainy neon street at night, low tracking shot following from behind, reflections on wet asphalt, handheld feel". Camera language such as tracking, dolly, orbit, or static wide changes the result more than extra adjectives do.
Describe one clear action per clip. If you need three beats, give them in order: "she opens the door, steps outside, looks up at the sky". When audio is on, mention the sounds you want, such as rain, footsteps, or distant traffic, so the generated track matches the scene.
For brand or character work, start from an image. Generate a still with Nano Banana 2.1 or GPT Image 2.5 in the same studio, send it to Seedance 2.5 as the first frame, and describe the motion. This two-step route gives you control over the look before you spend credits on motion.
In reference mode, use clean images with a single subject and consistent lighting. Label what each reference is for in the prompt. A mix of one face reference, one wardrobe reference, and one location reference works better than thirty near-identical portraits. First-frame mode and reference mode cannot be combined in one run, so pick the one that fits the shot.
Choose Seedance 2.5 for length and references. Choose Seedance 2.0 when 15 seconds is enough and you want a 4K option, or Seedance 2.0 Fast for quick drafts. Veo 3.1 suits short, cinematic clips with 1080p and 4K options, and MiniMax H3 is worth testing for fluid motion at 2K. Because every model is in the same studio, you can run the same idea through two engines and keep the better take.
Seedance 2.5 bills per second, and the studio shows the exact credit estimate before you generate. Draft at 480p and a short duration to test the idea, then re-run the keeper at 720p or 1080p and full length. Failed generations are refunded automatically, and paid plans include commercial usage rights for finished clips.
How to use Seedance 2.5
Subject, action, camera, and light. Mention sounds if Audio is on.
Who uses Seedance 2.5

Longer scenes with a consistent cast.

Vertical and widescreen cuts of one concept.

Product clips that keep packaging consistent.

Visuals with a steady art direction.
More models
Switch engines without switching tools — Nano Banana 2.1 plus the strongest public image and video models, side by side in the Nano Banana AI studio.
FAQ