Describe one scene, choose its duration, and create a short video.
Seedance 2.5
Red teapot: an actual generated example
A small red ceramic teapot on a cream table, subtle steam drifting upward, fixed camera, soft daylight, no lettering.
Seedance 2.5 is ByteDance's video model for building moving scenes from a written idea or supporting references. NanoPic gives you a place to plan the subject, motion, framing, and optional sound together. You can create a short visual study or allow a scene to develop for up to 30 seconds, but extra time is useful only when the action needs it. Begin with a four-second draft and a simple action. The red teapot preview is an actual Seedance 2.5 result from the displayed prompt: a quiet, low-resolution motion check, not a showcase of every feature.
Choose an input method that removes ambiguity without overloading the first attempt.
Describe who or what appears, what changes during the shot, and where the camera stays. A clear action such as steam rising or a curtain moving can communicate more than several competing events. Keep a beginning and an ending in mind even when the clip is only four seconds long.
Use the frame-guided workflow when the starting composition matters. An image gives the model a visual anchor, while your prompt explains the intended movement. If you supply a last frame, keep it compatible with the first. Large changes in pose, lighting, or perspective can make the transition harder to control.
Use relevant reference images to guide appearance and composition. The reference workflow accepts up to 9 images, subject to the upload limits shown in the controls. Describe movement and optional sound in your written prompt. Give each image a clear purpose and use only material you have permission to upload.
Build the scene in a short, deliberate review loop.
Use text-to-video for a fresh concept, image-to-video for a frame-led shot, or reference-to-video for supporting direction. Write one main action and one camera instruction. Keep the default teapot prompt if you want to explore a quiet scene before attempting people, fast movement, or a more complex environment.
Back to generator
The page starts at 480p, 4 seconds, landscape 16:9, with generated audio off. Choose the intended aspect ratio before generating. When the direction works, select 720p or 1080p, or a duration up to 30 seconds, if the project needs it. Check the displayed credit estimate after changing settings.
Back to generatorReview the opening, middle, and ending rather than judging a poster frame. Check subject shape, background stability, camera continuity, and unintended scene changes. If you enable audio, listen separately for timing and unwanted speech or music. Revise one major instruction at a time so the next attempt answers a clear question.
Back to generatorThese prompts are starting points to adapt. They are not claims that matching clips have already been produced.
Prompt: A small paper boat floating slowly across a shallow rain puddle, gentle ripples, reflected overcast sky, low fixed camera, one continuous shot, no lettering. Begin with a short silent draft. The scene asks for one readable movement and avoids a complex camera move that could compete with the subject.
Prompt: A closed sketchbook on a wooden desk, a breeze gently turns the cover open to a blank page, soft afternoon window light, slow restrained push-in, hold on the open book at the end, no lettering. Allow enough time for the action and the final hold, and check that the book does not change shape.
Prompt: A small lantern on a garden table at dusk; begin with the lantern unlit, its warm light gradually brightens, the surrounding leaves move gently, finish with a quiet hold, fixed camera, no cuts or text. Add duration only after a shorter version establishes the look. Avoid turning one continuous scene into an overloaded list of events.
Use generated motion to explore direction, then review it for the context where it will appear.
Create a brief mood shot for a journal, a fictional scene, or a creative project. Keep the visual message readable without sound if viewers may watch on mute. If the clip depicts a real person or event, distinguish an imagined illustration from genuine footage and obtain appropriate permissions.
Show collaborators how a camera move or a scene transition might feel before planning a larger production. Record the prompt, duration, and intended action with the selected draft. Treat the output as a direction reference; it does not guarantee that physical movement, spatial relationships, or production logistics are realistic.
Use a short environmental shot or object animation as one component in a larger edit. Choose vertical or landscape framing for the destination from the start. Add verified captions in your editor, check platform safe areas, and review looping behavior rather than assuming that the first and last frames will connect seamlessly.
Match the workflow to the information you already have and the result you need.
Text is useful when the scene is still open to exploration. A first frame is useful when the composition already exists. Neither method guarantees perfect subject preservation, so compare the output with the intended subject and watch for drifting details throughout the clip, not only at the start.
Use the video model hub to compare Seedance 2.5 with Wan 3.0, Kling O3, or MiniMax H3 on the same subject and motion brief. Review movement, continuity, available inputs, and the displayed cost. Different outputs or settings can change the comparison; one sample is not a universal ranking.
A four-second clip is enough to evaluate a simple movement and can keep early experiments focused. A longer duration gives an action room to develop but also creates more frames to inspect. Plan the sequence before extending it, and avoid spending more merely because a longer option is available.
Practical answers about inputs, examples, and preparing a usable result.
The generator supports 4–30 seconds. Start with the shortest duration that shows the action clearly. Longer clips need a readable progression and a full continuity review, not simply more descriptive words.
480p, 720p, and 1080p are available in this workspace. The default is a four-second 480p draft with audio off. This page does not offer a 4K generation setting.
Yes, through the appropriate frame or reference workflow. Reference-to-video accepts up to 9 images, with additional upload limits shown in the interface. Use relevant, permitted images and explain each one's purpose. Describe motion and optional sound in the prompt.
Generated audio is optional and starts switched off here. Enable it when sound is part of the scene, then review what was generated. Requested speech, ambience, or timing may need correction or replacement in a separate editing workflow.
The red teapot preview is an actual Seedance 2.5 video generated from its displayed prompt. It is a short, silent 480p example. The other prompt suggestions are ideas to adapt, not additional completed example videos.
This page creates videos; it does not provide region-level video editing. Generation consumes credits, and the current estimate appears before submission. Review account offers and download conditions rather than assuming unlimited free generation.
The page defaults to Seedance 2.5. Confirm that exact model name in the visible selector before submitting, especially after switching models. Reference uploads on this page are image-only; video and audio file uploads are not provided. The selected model and current controls describe the workflow that will actually run.
Start with a four-second draft. Keep the movement that communicates your idea and add complexity only when it serves the story.
