Not everything you need a video for is a product. Sometimes it is an announcement, an atmosphere, a piece of brand content that has no listing behind it. Freestyle mode is for that: you describe a scene in your own words and the AI agent plans and renders it.
How it differs from the product flow
The product flow starts from data, a URL, a title, images, and builds an advertisement around it. Freestyle starts from a description, and the agent expands that description into a full sequence: what happens, how it is framed, how it ends.
That makes it more open and slightly more demanding of you. With a product link the agent has facts to work from. With freestyle, the quality of what you get tracks the quality of what you asked for.
Describing a scene well
The difference between a vague result and a good one is almost entirely in the prompt, and the useful details are more concrete than people expect.
Weak: a nice coffee shop scene.
Better: early morning in a small wood-panelled coffee shop, low winter light through the window, steam rising from a freshly poured cup, the room otherwise empty, camera slowly moving in.
The second version specifies time of day, light, material, subject, and camera behaviour. Each of those is a decision the AI would otherwise make for you, probably differently from what you pictured.
Three things to include if you include nothing else: the light (time of day, indoor or outdoor, warm or cold), the subject and what it is doing, and what the camera does.
Settings
A settings card lets you choose the model, the duration, the aspect ratio and whether to add a spoken voice-over. The cost of the generation is shown before you confirm, so nothing is charged on a surprise.
If you want a specific look rather than a described one, you can upload an image to guide the image-based models. This is the most reliable way to keep a consistent visual identity across several freestyle videos.
Voice-over
The agent writes the spoken lines in the language you are chatting in, so the narration matches the audience rather than sounding translated. It is optional; plenty of social content works better silent with on-screen text.
Where freestyle earns its place
- Brand and lifestyle content with no single product at the centre.
- Announcements, openings, seasonal messages, anything with a date rather than a SKU.
- Concept testing before committing to a full campaign, when you want to see whether an idea reads before anyone spends on it.
- Intros and transitions for longer content you assemble elsewhere.
What you get
A standard MP4 in 9:16, 16:9 or 1:1, at 720p by default and 1080p on the premium models. Longer pieces, up to 36 seconds, are built as multiple connected chapters of 12 seconds each with audio continuity between them, so a longer video develops rather than repeats.
Generating variations is worth the habit here more than anywhere else, because with an open brief the first result is a sample of what the description allows, not the only thing it allows.
Related reading: How the AI agent writes your script and Meet the AI agent.