Canvas creation and node workflows

Open canvas creation

A canvas brings prompts, images, videos, and audio into one workspace. Connect nodes to reuse inputs across your workflow. Production canvases belong to a project and can also belong to an episode. Canvas content is saved to the cloud, and generation uses real models with the corresponding credit charges.

Before you start

• Sign in, confirm access to your target project, and make sure canvas access is enabled for your account.
• Viewing and editing require different permissions. A read-only canvas can be viewed; adding nodes, uploading media, connecting nodes, and generating require canvas write access.
• Editing playlists also requires editor write access. Rendering a finished video requires a valid project and episode, plus available cloud export capability for that project.
• Check the model, estimated cost, project quota, and team balance before generating. Reference limits, video duration, and available settings depend on the current model.

1. Open the correct canvas

There are two common routes:

1. On the canvas creation page, choose the new canvas action, enter a project name, and create the project and open its canvas. This creates a project in the team you own and opens its default canvas. The create action requires an available owned team.
2. For an existing project, open it from the project list, enter the target episode, and choose the open canvas action. The first visit creates that episode's default canvas when needed; subsequent visits open the same canvas.

The project workspace lists existing canvases and lets you open them, rename them, or archive them when you have write access. The canvas creation page lists canvases you have recently opened so you can resume work.

Check the project, episode, and canvas name in the header. A canvas at project level can organize media and generate content. Open an episode canvas when you need playlist cloud export.

2. Add prompts and media

1. Use the create action in the bottom toolbar to add a prompt, image, or video node. You can also open the creation menu by right-clicking an empty part of the canvas.
2. Enter a prompt in an image or video node, or write it in a separate prompt node to reuse it across several generation nodes.
3. Upload images, videos, or audio and wait for the upload to finish before using them. The page reports unsupported file types and size limits.
4. Open the media picker to reuse project assets. On an episode canvas, switch between the current episode and the current project, then filter by category or search by name.

Uploaded or existing audio can be used in workflows. Audio generation on the canvas is currently unavailable.

3. Connect nodes and generate

1. Drag from the output port on the right of an upstream node to the input port on the left of the target node.
2. If an input-purpose menu appears, choose the prompt, reference, or frame input you need. Unavailable choices show their restriction.
3. Select a model in the target image or video node and review the available aspect ratio, resolution, duration, and other settings.
4. Creative mode uses a prompt and supported references. Frame mode requires at least a valid first-frame image and a model that supports this mode.
5. Review the estimated cost in the node, then generate. Relevant inputs are locked while a generation or recovery query is active; wait for the operation to finish before changing them.

When a prompt node is connected, the target uses its prompt content. Recheck your inputs after changing the model or mode. Frame mode cannot directly reuse the material mentions from creative mode.

Nodes show processing, success, or failure states. After success, preview or download the media and use the node details to inspect result history and select a result for further work. Adding content to an upstream node does not automatically submit downstream generation; confirm and submit on the target node.

4. Arrange clips in a playlist

With both canvas and editor write access, add a playlist node and open its playlist editor.

1. Add images, videos, or audio from the canvas, or import additional media with the local upload or asset library actions in the playlist.
2. Add the media you need to the timeline, adjust clip order and boundaries, and use split and trim actions to arrange the content.
3. Review playback, audio tracks, volume, and output aspect ratio. Confirm that every source can be read.
4. Open the timeline export menu and choose to export and add the video to the canvas, or export and download an MP4.
5. Read the estimated credits and duration in the quote confirmation, confirm the export, and wait for server rendering to finish.

Both finished-video export actions use cloud rendering. Downloading an MP4 from a playlist also requires cloud export permission and a quote confirmation. Cloud export currently supports up to 10 minutes. The playlist cannot be edited during export. After you change the timeline, export again to produce a video that reflects the new content.

Exporting and adding to the canvas inserts the finished video as a video result. To save an individual source video, use the node download action or the playlist's original-video download action without rendering the entire timeline.

For more complete track, subtitle, and effect editing, see Video editing and finished-video export and open the editor from your project episode.

5. Confirm saving before leaving

Changes to nodes, connections, and layout are saved automatically. The header shows pending, saving, or saved status. Confirm that saving has finished before leaving. If saving fails, retry from the header and keep the page open until it succeeds.

If you leave with unsaved changes using the back action, the page asks you to save and leave, discard changes, or continue editing. Discarding leaves those changes out of the cloud version.

Use canvas and node names to identify stages of your work. Color marks, the header filter, and node search help you navigate larger workflows. Undo and redo apply to normal editing; clearing the canvas removes its nodes and connections and is explicitly marked as irreversible in the confirmation dialog.

Save conflicts and task recovery

• Version conflict: If another page or device updates the same canvas, automatic saving pauses. Read the notice, check whether a local recovery copy was saved, and then reload the server version.
• Local recovery copy found: After reopening, inspect the cloud content and choose whether to restore the local version or discard the local copy. Restoring replaces the current server canvas; review the confirmation carefully.
• Recovery copy could not be saved: Keep the page open. Your current changes may exist only in the page. A recovery copy belongs to the current browser and does not mean the cloud content is saved.
• Generation still active after refreshing: The system attempts to resume status queries for existing tasks. If the node offers a continue-query action, use it first. Check the original task before submitting a new generation.
• Unavailable media or failed generation: Follow the node message and check permissions, project status, the model, references, and credits. Reselect or upload media that cannot be read.
• Playlist export unavailable: Check the episode association, canvas and editor write access, and cloud export availability. Use the recheck cloud export action to retry the capability query.

What you will have afterwards

Your canvas keeps its nodes, connections, media references, and generation results. Reopen it from recent canvases or the project workspace to continue. A playlist export can be reused as a canvas video result or downloaded locally.

Project workspace and episodes

Video editing and finished-video export

Back to the user guide