Choose a generation mode
Learn when Agent generates images, video, and audio directly on the canvas, and when it waits for your approval first.
In Creative OS, you can tell Agent, “Create two 16:9 product images,” “Turn this image into a five-second video,” or “Make background music for this scene.” Agent chooses an image, video, or audio model for the task, prepares settings such as aspect ratio, duration, and output count, then calls the model and places the result on the canvas.
Before generation begins, you can choose how Agent should proceed:
- Auto: Agent starts generating as soon as the model and settings are ready. You do not need to approve each generation.
- Ask: Agent shows a confirmation card first. You can review or change the model and settings. Agent starts only after you click Generate.
Use Auto when you want Agent to continue through a longer workflow. Use Ask when you want to check the model, aspect ratio, duration, and output count before each generation.
Both modes use Tapies whenever a model actually generates an image, video, or audio result. The amount is based on the selected model and settings. Ask mode only delays that spend until you click Generate; it does not make the generation free.
Auto: Agent continues directly
In Auto mode, Agent chooses models and settings from your task, canvas references, and current context, then starts generating immediately. New images, video, or audio continue to appear on the canvas without a separate approval step for each generation.
Use it for longer workflows with a clear goal. For example, Agent can organize a brief, draft a script, generate storyboards, and continue making assets without needing you to adjust settings after every step.
Before you start, make sure that:
- the task scope is clear;
- product images, character references, and documents are the right version;
- the output count and specifications are what you expect.
Best for a proven workflow
Auto removes waiting between steps, so model generation and Tapies spend begin sooner. Use Ask mode first when you are testing a new direction or the references are still changing.
Ask: review a confirmation card before generating
In Ask mode, Agent does not call a model right away. It first creates a confirmation card with the settings it is ready to use.
In the card, you can review or change:
- the model;
- image aspect ratio, video duration, and other generation settings;
- output count;
- the images, video, audio, or files being used as references.
When everything looks right, click Generate. Only then does Agent call the model and create the result on the canvas. If something needs to change, update the card or add a note in the conversation, then confirm again.
Ask mode is useful for a first attempt, work that needs exact product or character consistency, long video or batch generation, and any task where you want to inspect the model and settings before running it.
Which one should you choose?
| Your situation | Recommended mode |
|---|---|
| A proven workflow with several steps for Agent to complete | Auto |
| A new model, new reference, or new direction | Ask |
| You need to adjust the model, ratio, duration, or output count first | Ask |
| Specifications are fixed and you need a quick set of similar assets | Auto |
You can switch modes at any time. A practical habit is to use Ask mode to establish a workflow, then switch stable work to Auto mode.
Next: Manage conversations