Video generation
Short brand videos from a brief or a starting image, made with several video models, with each clip's cost shown before you generate and billing per second from your video pool.
Video generation is the second media pillar: short, on-brand clips generated in-house from a plain-language brief or a starting image, in the social composer and in Media Studio, and saved to your Media Drive alongside your images. It runs on the same pipeline as image generation, so requests, progress, and delivery work the same way.
What it does
The video studio turns a brief into a short brand clip:
- Text-to-video, describe the shot and AnyBot refines it into one brand-aware prompt, then the video model you pick renders a short clip.
- Image-to-video, steer the look and motion from a starting image, so a clip can grow out of a still you already generated, or turn a brand image into a short animated video.
- More control, add reference images, set a last frame, start from an existing clip (video-to-video), and use a negative prompt to say what to leave out.
- Short clips, around 5–10 seconds per generation, in standard or pro quality.
- Ratios, 16:9, 9:16, and 1:1, so a clip fits a feed, a story, or a square post.
- On-brand by default, the same brand-aware refinement as images, drawing on your Materials & Brand Kit.
Several video models
Video doesn't run on a single model. One video interface fronts several video models, so new models can be added without changing how you generate. You pick a model in the studio, and the cost of the clip is shown before you generate.
Because they share one interface, differences between models (supported durations, whether a clip can start from a reference image, output resolution) are read from each model's capability profile rather than hard-coded, and the studio adapts the options it shows you to the model you pick.
Same pipeline, same billing
Video rides the shared async flow, so there's nothing new to learn:
- You send a brief and get an immediate "pending" acknowledgement, the request never blocks.
- The media worker refines your prompt, renders the clip, and uploads it, streaming live progress to your Tasks Center.
- The finished clip lands in your Media Drive and the usage is metered.
For the full lifecycle, states, realtime progress, and why a retry never double-charges, see How generation works.
Video is metered per second on its own dedicated pool, separate from images and from avatar/presenter, so one media type never eats another's quota. A failed generation isn't charged, and a retry only bills if it succeeds. See Media billing for the dimensions.
FAQ
Can I generate video today? Yes. It draws from your video pool, billed per second of output, and every paid plan funds that pool.
Which video model runs? You choose from several video models in the studio. The options shown (length, ratio, quality) follow the model you pick, and the clip's cost appears before you generate.
How long are the clips? Around 5–10 seconds per generation, depending on the model.
How is video billed? Per second, on a dedicated video pool, see Media billing.