Chat with Agent

Open Agent, send a task, reference canvas content, upload files, and save your usual generation preferences.

Open Agent on the canvas, describe what you want done, and add the images or files it should use. There is no required prompt formula. Start with the action, then add the references, output settings, and any point where Agent should wait for your approval.

1. Open Agent

You can open or close Agent in two ways:

  • Press ⌘ J on a Mac, or Ctrl J on Windows.
  • Click the Agent icon in the bottom-right corner of the canvas.

Once Agent is open, click the input at the bottom to begin.

Open TapNow Agent

2. Type and send

Write the task in plain language. When you are ready:

  • Press Enter to send.
  • Press Shift + Enter to start a new line.
  • Or click the send button on the right side of the input.

An instruction is easier to execute when it includes four things:

  1. Action: what Agent should do.
  2. References: which image or file it should use.
  3. Output: the quantity, aspect ratio, duration, or format.
  4. Boundaries: what must stay unchanged and whether Agent should wait for approval.

Copy this instruction and try it with TapNow Agent

Create two 16:9 camping ads for a portable outdoor speaker. Set the scene at a campsite at night with two people. Place the speaker naturally on the table. Preserve the product design and logo, and generate two images now.

You do not need to put every detail in the first message. After Agent responds, continue with a focused correction such as, “Keep the composition, but reduce the group to two people,” or, “The second direction works better. Make two variations of it.”

If you only want an assessment, say, “Analyze this first. Do not generate anything.” If you are ready to make something, specify the output and whether Agent should wait for approval.

3. Reference images on the canvas

There are two ways to reference canvas images.

Lightweight reference: add images to the current context

Click the + beside the input and choose Insert from canvas. Images that are already selected will be added first, and you can continue clicking other images on the canvas. Added items appear above the input.

Use this when you simply want Agent to consider a group of images in the current task. Once they are in the context, you do not need to name every image in the instruction.

Select images from the canvas

Explicit reference: use @ in the instruction

When images have different roles, add them to the context first. Then type @ in the input and choose a specific image. This lets you say which image defines the product and which one is only a style reference.

Reference an image with @ in the instruction

Copy this instruction and try it with TapNow Agent

Use @Front Product Image to preserve the package structure, logo, and text. Use @Camping Reference only for the nighttime lighting and color. Create two 16:9 ads, and do not copy the people from the reference.

Before sending, check the references above the input. Add anything that is missing and remove older references that do not belong in the task.

4. Upload files

When the source material is not on the canvas, click the + beside the input, choose Upload attachment, and select a file from your computer. Uploaded files appear above the input.

Tell Agent what each file is for instead of sending an attachment by itself.

Upload a file as an attachment

Copy this instruction and try it with TapNow Agent

Read the client brief and product image I just uploaded. First list the required selling points, delivery specifications, and open questions. Then draft a four-shot script. Do not generate images or video yet.

For a long document, start with a short summary. Once you know Agent has the correct version, continue with the production task.

5. Save your usual preferences

For settings you use regularly, tell Agent, “Remember this.” Agent can carry those preferences into later conversations, including your default model, aspect ratio, output count, and approval flow.

A task-specific instruction always takes priority. You can override a default whenever a job needs something different.

Copy this instruction and try it with TapNow Agent

Remember my default image settings: use Seedream 5.0, a 16:9 aspect ratio, and generate two images each time. If I specify a different model, aspect ratio, or quantity for a task, follow the instructions for that task.

Copy this instruction and try it with TapNow Agent

Remember my default video workflow: use a 9:16 aspect ratio. Show me the shot plan and model choice first, then wait for my approval before generating.

Copy this instruction and try it with TapNow Agent

Remember this rule for product work: when a task includes a product reference image, preserve its package structure, logo, and text. Ask me if anything is unclear.

When a preference changes, say, “Change my default aspect ratio to 1:1,” or, “From now on, generate directly without asking me to approve the shot plan.”

Quick check before sending

  • Agent knows what it should complete.
  • Every required image and attachment appears above the input.
  • The quantity, aspect ratio, duration, or other output settings are clear.
  • Fixed details and approval points are unambiguous.

Next: Choose a generation mode