Complete Canvas Tutorial
The Canvas is not just a file preview area — it's a visual creative workstation designed around assets, Prompts, reference images, video generation, and iterative workflows.
Canvas Asset Board & Timeline Editor
From idea to finished video, all in one workspace
Traditional AI video workflows require switching between multiple tools — generate, download, organize, edit. Canvas integrates asset management, AI generation, and video editing into a single desktop interface, letting creators focus on their ideas.
Asset Board: Your AI Creative Hub
Images, videos, audio, and text in your workspace are automatically displayed as cards on the canvas. Drag in local files and they appear instantly; AI-generated content lands in place automatically — no manual organization, no lost assets.
Submit and a countdown appears immediately; thumbnails update automatically on completion. Failed generations don't require re-filling parameters — retry with one click.
Select a text asset and call Gemini or Claude to optimize scripts, dialogue, and descriptions in one click.
Infinite pan and zoom; nodes are freely draggable. Supports Cmd+Z undo — your creative process is always reversible.
Multi-model AI generation, parameters at your fingertips
Built-in leading generation models — generate images and videos directly from the input bar:
- Image: Nano Banana 2 / Nano Banana Pro / Midjourney V7, supporting 1k~4k resolution
- Video: Seedance 2.0 / Seedance 2.0 Fast, 5~15 seconds freely adjustable
Two video generation strategies
- Reference mode: @ reference images/videos on canvas as style or content guidance; AI understands your intent
- First/last frame mode: specify the start and end frame; AI fills in the motion in between
Timeline Editor: Lightweight editing, fast output
Desktop only — not available on WebBrowse workspace files in the left asset library (list/grid view) and drag them directly into the video or audio track. The preview area follows the playhead in real time — what you see is what you get.
Space to play/pause, arrow keys for frames, Cmd+B to split, Delete to remove — edit without leaving the keyboard.
Canvas and timeline share the same workspace. AI-generated assets appear on the canvas automatically and can be dragged into the timeline at any time. From inspiration to rough cut — no export, no tool switching, one window.
Dual-track editing
- Video track: clips snap into alignment automatically; drag to reorder
- Audio track: freely position clips to match the visual rhythm
1. What is the Canvas#
Canvas automatically scans asset files in the current Workspace and displays them as nodes.
Supported asset types include:
- Images
- Videos
- Audio
- Text files
- Markdown files
- In-progress AI generation task nodes
- Completed AI generation result nodes
Any file saved to the current Workspace directory will automatically appear on the canvas.
2. Entering the Canvas#
Open a Workspace and switch to the Canvas view. If the Workspace already contains images, videos, or audio files, they will appear on the canvas automatically.
If the canvas is empty, you can:
- Drag in local files
- Generate new images or videos from the bottom input bar
- Copy files into the Workspace folder
3. Basic Canvas Operations#
| Pan the canvas view | Drag on an empty area of the canvas to move the entire view. |
| Zoom the canvas | Use the mouse scroll wheel, trackpad, or the canvas control buttons to zoom. |
| Move asset nodes | Drag any node to reposition it. The canvas remembers node positions; your layout is preserved when you reopen the Workspace. |
Select nodes
- Click: select a node
- Click again: deselect
- Shift + click: multi-select
- Drag to box-select: batch select
- Click empty area: clear selection
Delete nodes
Select one or more nodes and press Delete or Backspace. A confirmation dialog will appear.
Undo and redo
- Cmd / Ctrl + Z: undo
- Cmd / Ctrl + Shift + Z: redo
4. Adding Assets to the Canvas#
Method 1: Drag in files
Drag local files directly onto the canvas. Files are copied into the current Workspace and automatically displayed as nodes.
Method 2: Copy to the Workspace folder
Manually copy asset files into the Workspace directory. The canvas will scan and display them automatically.
Method 3: Generate via the canvas
Use the bottom generation bar to generate images or videos. New assets appear on the canvas automatically after generation.
5. Using the Bottom Generation Bar#
The input bar at the bottom of the canvas is the main generation entry point. It contains: generation type, model selection, reference mode, aspect ratio, resolution, video duration, Prompt input, and the send button.
5.1 Select generation type
- Image: generate images
- Video: generate videos
5.2 Enter a Prompt
Prompts work best in English. Even if the interface is in another language, write generation prompts in English.
5.3 Set image resolution
- Draft stage: use 1k
- After selecting a direction: use 2k
- Final image or poster: use 4k
5.4 Set video duration
- Testing motion: 5 seconds
- A complete single shot: 8~10 seconds
- Complex motion or narrative change: 12~15 seconds
6. Image Generation Tutorial#
6.1 Simplest image generation
Select Image → choose image model → set ratio and resolution → enter Prompt → click send
6.2 Recommended Prompt structure for images
Write in this order: subject description + style/texture + lighting/tone + ratio/composition
6.3 Image-to-image
If you already have a reference image, use @ in the Prompt to reference images on the canvas.
7. Video Generation Tutorial#
7.1 Simplest video generation
Select Video → choose Seedance model → set ratio → set duration → enter Prompt → click send
7.2 Video Prompts must include motion
Video is not a static image. Your Prompt should explicitly describe: how the camera moves, how the subject moves, whether the background has motion, and whether lighting or the environment changes.
7.3 Common camera movement terms
8. Using @ to Reference Assets#
Type @ in the Prompt input box to select assets from the Workspace.
8.1 Uses for asset references
- Maintain character consistency
- Reference existing scenes
- Reference visual style
- Reference video motion
- Reference audio rhythm
- Inherit characteristics from existing assets
8.2 Reference limits per model
If too many assets are referenced, or if the referenced type is unsupported by the model, the input bar will show an error and the send button will be disabled.
9. First/Last Frame Video Generation#
Video mode supports two reference methods: Refs (standard reference) and Frames (first/last frame).
| 9.1 Refs mode | Character reference / Scene reference / Style reference / Video motion reference / Audio rhythm reference |
| 9.2 Frames mode | Use this to control exactly which frame the video starts and ends on. Set First Frame and Last Frame. |
9.3 Ideal use cases for first/last frame
- Day to night
- Character turns from front to side
- Product opens or closes
- Camera pushes from wide shot to close-up
- Magical transformation
- Scene transition
- Emotional shift
10. Viewing Asset Details#
Double-click any node to open the details panel, showing: file name, path, type, generation model, ratio, resolution, duration, original Prompt, and preview.
Great for reviewing past versions, comparing results, and recalling Prompts you used before.
11. Regenerating from Existing Assets#
Open an image or video details panel and use the regeneration area on the right or bottom.
- Fine-tune the image
- Keep the character, change the scene
- Keep the scene, change the lighting
- Turn an image into a video
- Explore new versions based on an existing video
13. Text Optimization#
If a node is a text or Markdown file, the details panel displays the content with a text optimization entry (Gemini or Claude). Useful for: Prompts, storyboards, character profiles, scene descriptions, ad copy, and video scripts.
14. Recommended Creative Workflows#
14.2 Product advertisement
Import product image → generate stylized background → use product image as strong reference → generate multiple ad visuals → select as first frame → generate product animation → compare multiple versions
14.3 Storyboard exploration
Generate key shots in image mode → arrange in order on canvas → write video Prompt for each shot → generate in video mode → use approved videos for editing
15. Prompt Templates#
Image: [subject] + [action/state] + [style] + [lighting] + [ratio/composition]
Character consistency: @character-ref, [character description], [action], [scene], keep appearance consistent
Scene consistency: @scene-ref, [new subject], preserve the same environment and atmosphere
Video: [subject] + [motion description] + [camera movement] + [environment] + [style]
First/last frame: Starting with [first frame], ending with [last frame], [subject motion], [camera movement]
16. FAQ#
17. Advanced Tips#
Don't jump straight into complex video generation. First stabilize your character, scene, style, and key frames — then generate video.
Left: characters. Top: scenes. Right: video results. Bottom: discards and alternates.
A well-structured Workspace makes @ references more precise and reduces mistakes.
Keep multiple versions (e.g. character_v1.jpg, character_v2.jpg) for comparison and rollback.
Clear face, clear lighting, no complex occlusion, subject fills most of the frame, no multiple similar figures in one image.
Keep character references and background references separate — don't use one image for both.
For a 5~10 second video, don't pack in too many events. Describe one main action per generation.
Use Frames mode when you need a clear start and end point. Use Refs mode when you're just providing style or character reference.
Reduce referenced assets, shorten the Prompt, keep only one main action, and test with the Fast model first.
For example, change only lighting, camera, character action, background, or style at a time. This helps you pinpoint what made the result better or worse.
Best Practice Summary
The clearer your assets, the more explicit your references, the more specific your Prompt — the more consistent your results.