Prompt-led direction
Describe the subject, action, camera movement, pacing, and atmosphere in one focused prompt.
Seedance 2.5 text-to-video
Create a cinematic video up to 30 seconds long from a prompt, first and last frames, or multimodal reference media with Seedance 2.5.
Output
Your video will appear here
Add a prompt and optional reference media, then choose your output settings.
The model, explained
Seedance 2.5 is ByteDance's next-generation cinematic video model. It creates coherent shots up to 30 seconds long while following the subjects, action, camera direction, and atmosphere in your prompt.
Choose first and last frames for a controlled transition, or use up to 50 image, video, and audio references to guide characters, products, motion, timing, and sound.
FramePack brings the Seedance 2.5 workflow into one place with secure uploads, 480p and 720p output, synchronized audio, asynchronous progress, and downloadable results.
Built for cinematic video
A longer, more controllable workflow for turning ideas and multimodal references into directed AI video.
Describe the subject, action, camera movement, pacing, and atmosphere in one focused prompt.
Upload up to 30 images, 10 videos, and 10 audio files to guide identity, timing, motion, and sound.
Enable native synchronized audio when the scene needs sound that follows the generated action.
Choose a 4–30 second duration, seven aspect ratios, 480p or 720p output, and optional web search.
Start with Seedance 2.5, or switch to the retained 2.0, Fast, and Mini models for different quality, speed, and cost tradeoffs.
Every request is sent with safety checking enabled for a safer generation workflow.
Three simple steps
Start with a clear idea, add only the references that help, and let Seedance build the scene.
Write the subject, action, camera direction, pacing, and visual style you want to see.
Add first and last frames, or upload image, video, and audio references to guide the result.
Select the aspect ratio, resolution, duration, audio, and search settings. Submit, then preview and download the finished video.
Specific prompts and purposeful references help the model keep the generated movement coherent.
Questions, answered
Seedance 2.5 turns a text prompt and optional frame or multimodal references into a cinematic video up to 30 seconds long.
Your settings and uploaded references are validated by our server and submitted as an asynchronous generation task. The page follows the task until the result is ready.
First and last frames are mutually exclusive with reference images, video, and audio. Choose the workflow that best matches the control you need for the shot.
Generation time varies with duration, resolution, generation-service capacity, and queue length. The output panel shows the latest task progress while FramePack processes the video.
Write a focused prompt, upload the references that matter, and create a cinematic Seedance video online.
Create with Seedance 2.5