Limited offerAnnual membership on sale — unlock more image & video generation
Resources›Canvas›Infinite Canvas basics: from a blank canvas to your first generation chain

Infinite Canvas basics: from a blank canvas to your first generation chain

Connect text, image, video and audio nodes on one canvas, feed upstream results into downstream nodes, restore any version and share in one click.

CanvasBeginner8 min6 steps
Step 1

Create a project and a canvas

Open Infinite Canvas in the sidebar and click New project. The first canvas is created and opened automatically.

A project can hold several canvases. Deleted projects go to the Trash under My projects and can be restored.

Step 2

Add nodes

Click + at the top-left of the canvas or right-click on empty space to add one of four node types:

  • Text: write prompts, or let AI expand / rewrite them.
  • Image: text-to-image, image-to-image, multi-image, face swap, motion control.
  • Video: text-to-video, image-to-video, keyframes, omni reference, motion control, lip sync, video face swap, upscale, edit / extend.
  • Audio: text-to-speech.

Nodes can be dragged, box-selected and duplicated; the top-right toolbar has alignment and zoom.

Step 3

Write a prompt, pick a model and a mode

Select an image or video node and the composer appears at the bottom:

  1. Pick a model (e.g. Flux / Qwen for images, Seedance / Wan for video).
  2. Pick a mode chip. Each mode has its own material slots: face swap needs face + target image, keyframes need first + last frame; hints are shown next to the chip.
  3. Write the prompt. Modes such as face swap or upscale need no prompt and hide the input.
  4. Choose ratio / resolution / duration; the credits required update live at the bottom-right.
Step 4

Connect nodes: reference upstream results

Drag the dot on the right of an upstream node onto the left of a downstream node to create a reference:

  • Image → Image: reference for image-to-image / face swap / multi-image.
  • Image → Video: first frame, last frame or reference image.
  • Video → Video: source video for face swap, upscale, edit or extend.
  • Audio → Video: reference audio for lip sync.

When the upstream node is regenerated, the downstream node shows Reference updated; click it to re-run with the new material. Slot limits depend on the mode and are checked before submission.

Step 5

Generate, track progress, view results

Click Generate and the node switches to running with a live progress bar. Refreshing or reopening the page resumes polling automatically.

Results appear on the node when done: preview, download, or drag them into the Library. If a task fails, the node shows the reason and credits are refunded automatically.

One account can run only one task at a time; wait for it to finish before submitting the next.

Step 6

Versions, sharing and export

  • Versions: every completed generation takes a snapshot; History (top-right) restores any version, and restoring creates a new version instead of overwriting.
  • Share: creates a read-only link valid for 7 days; viewers need no account, and you can revoke it any time.
  • Duplicate canvas: copy the whole canvas with results and keep editing.
  • Export: download all results on the canvas as one package.