Canvases are easy to work with and a powerhouse tool for scaling your creative work. As you're getting started, here a few tips:

People with access to the same canvas may be able to view it, but only the creator can edit it. Canvas does not support simultaneous co-editing on a single canvas.

On editable canvases, access direct editing tools from the left-side rail or the selected-image toolbar's Tools menu. If nothing is selected, choose a rail tool, then an image. Use Crop mode for batch framing and extension, or tools like Smart resize, Camera angle, Convert, and Adjust image. You can also open Canvas in fullscreen for more working area, with fullscreen controls available in the toolbar.

Air Canvas can use supported images, videos, and audio as references for generation and editing, subject to model compatibility. Even for requests such as "make this brighter," generative workflows can alter elements beyond those requested, including text and logos. Generative workflows do not guarantee unchanged source footage or source quality.

Images already on your canvas appear as thumbnail references when you're chatting with Air. You can click to select them, or reference them directly by saying "this image" or "the poster on the left."

Air will understand which image you mean and apply the requested change.

You can also type @ in the Canvas AI chat composer to browse suggested canvas elements. As you move through the suggestions, the matching element is highlighted on the canvas with a dashed border and label.

After a video is processed, select it on the canvas or reference it in Canvas agent, then describe the style or content edit you want. Auto routes the request to a compatible model. Select a model manually when you need more control and the model supports the provided reference.

Using a video as a reference in Canvas is different from trimming, splitting, or extracting sub-clips from source footage in the asset viewer. For those workflows, see Editing and sub-clipping videos in Air.

Right-click an empty or populated canvas and choose Add from board to open the floating asset picker. From the picker, you can browse boards, search your workspace, choose a Context source, and select multiple assets to add to the canvas.
​
​Picker behavior: The picker closes after the selected assets are successfully added. If you drag an asset somewhere that cannot accept it, the picker stays open so you can try again.

You can drag and drop images directly into the chat, or upload them through the interface. Once uploaded and processed, Air can begin working with them immediately.

You can paste a direct image URL (like https://example.com/photo.jpg) and Air will load it from there.

You can save a web image directly into a Canvas with Air Web Clipper. Right-click the image in your browser, choose Save to Air, then select the Canvas destination. For setup and step-by-step clipping instructions, see Explore Air Web Clipper for Chrome.

Air works with all common image formats:

When Air delivers final images back to you, they'll typically be:

If you need to work with a specific area or object, you can describe it.

Give these sample prompts a try:

Air can analyze images to locate specific objects when needed.

When you have one image: Just say "this image" or describe what you want done. Here are some example phrases to use in your prompts:

You can also use @ mentions to identify a specific canvas element before you submit your request. While a suggestion is selected, Canvas previews the referenced element directly on the canvas.

For a deeper walkthrough of image-editing workflows, see Editing content in Canvas, which covers workflows like Crop & Extend, Smart resize, Change lighting, Camera angle, and Convert.

Stop an agent response
While a Canvas agent response is in progress, select the stop control in the chat composer. Canvas returns your submitted prompt to the input so you can edit or resubmit it, and no partial response remains in the chat history.
​
If you select a model before sending the first message in a new Canvas, Canvas uses it for the first run and saves the selection with the Canvas once it exists. For model guidance, see Editing content in Canvas.

Canvas agent chat history retains the prompts used in that Canvas. If you need the original prompt for an output, reopen the Canvas agent chat for that Canvas instead of checking asset metadata or image details.

Bookmark a useful user message. Create a reusable Skill from the post-run Save skill suggestion or action, which prefills its fields, or choose + Add in the slash-command picker to configure one manually. Skills save prompt-based instructions, not attached assets, generated outputs, or the full conversation. For visibility, permissions, and configuration, see Explore Skills.

For long-running Canvas agent sessions, Canvas may display a message when the thread is too long. Select Start a new thread, then bring over any prompt context you still need from the previous chat history.

When you have multiple images: Be specific about which one you're referring to. Here are some example phrases to try out in your prompts::

You can also click to select specific images before making your request.

The same reference behavior also applies to video generation workflows. If your canvas includes image, video, and audio references, you can combine them when prompting Canvas.

Air can store and use your brand assets — logos, fonts, and colors — so you don't have to hunt them down every time.

Your workspace supports multiple Context sources for brands, campaigns, channels, or audiences. Any workspace user can view them. Workspace admins can also create, modify, or delete Context sources, including asking the agent in Canvas to create or regenerate Brand Kit information from a website URL.

If your team has set up a Brand Kit, you can reference logos by name.

Examples:

For detailed setup steps, see Setting up context sources.

The same approach for adding in images to your canvas also works for fonts.

Give these sample prompts a try:

If the specific font isn't in your Brand Kit yet, Air can help locate it or allow you to upload it.

When brand colors are saved to your brand kit, simply reference them by name.

Give these sample prompts a try:

You can also give hex codes directly: