AI storyboard and short drama planner
Develop a short drama as a connected script, characters, scenes and ordered shots. Preview blocking with editable 3D cameras and keyframes, generate native H3 video versions, and share a fixed workflow for others to copy into private projects. Choose H3, Seedance 2.0 or Seedance 2.5, with audio and subtitle controls.
Open toolUpdated
- What you provide
- A story brief, character and scene details, 5/10/15-second shot timing and optional private image references.
- What you get
- A saved storyboard, 3D previews, private H3 video versions and reusable workflow JSON or revocable share links. Ordered previews and a private MP4 film assembled from selected video versions.
How to use it
- Describe your story and choose a video model. The Agent creates character, look, prop, scene, whitebox and frame cards, then automatically generates missing first images in dependency order. A selection limits the editing scope.
- Adjust 3D staging, cameras and keyframes. H3 shots use 5/10/15 seconds; Seedance 2.0 uses 4–15 seconds and Seedance 2.5 uses 4–30 seconds.
- Choose a model, describe the shot and add pictures in order. Review the settings and quote, then generate a new shot version. Prompt optimization is optional and must be triggered manually.
- Open Film to see shot readiness and total duration, play shots in order, then create and download an MP4 from the selected versions.
- Use Workflow to share a fixed saved version or download JSON. Others can preview it and create their own private copy.
- Choose H3, Seedance 2.0 or Seedance 2.5, with audio and subtitle controls.
Common questions
Can it generate videos on the site?
Yes. Review the quote to generate H3 clips of 5, 10 or 15 seconds at 768P or 2K. The Agent prepares shot settings; it does not automatically submit paid video generation. Completed shots show their video poster and player directly on the canvas. Choosing another version updates the card to that video.
What does workflow sharing include?
A fixed version of the script, node graph, whitebox animation and video settings. The link can be revoked, and copying creates a separate private project without the original media or history.
Do I need to generate every material image manually?
The Agent generates missing first images when creating or filling a project. Existing images and tasks are reused; retry failures from the cards. Shots keep their dependencies and wait for you to start video generation.
Can missing images generate in parallel?
Yes. Fill missing assets in parallel sends ready cards to the existing image queue. Each completed image unlocks its dependents without waiting for unrelated branches. Actual concurrency depends on queue and model limits. Failures block only dependent branches. Keep the workspace open; stopping prevents further submissions.
How do connected shots keep continuity?
The shot plan distinguishes continuous action from cuts. Connected shots show a tail-frame link. The server extracts the preceding video’s final frame, using its latest version by default or a version you select. An unfinished latest version must finish first. Cuts can stay independent. Set parameters and review the quote in the sidebar, then generate directly.