A continuous sequence is easier to control when each shot is an asset and each seam has an explicit image. The canvas then shows the source, the frame and the next clip as one inspectable chain.

Plan the shots

Write one sentence per shot: its subject, camera movement and end state. Generate or upload the first image, then choose a video model from list_models whose reference slots fit the shot.

Quote duration, resolution and references with estimate_credits before creating a clip. Video is priced by its configured request, so a small settings change can change the total materially.

Lock a seam

When a clip is ready, create an image with the free image/frame model. Use the clip in its source reference slot and select last or a specific time. That frame becomes the start-frame reference for the next clip.

Keep movement continuous across the cut: match direction, lens, subject position and lighting, then describe only the next action. If a model accepts an end frame, use it when the destination composition matters more than open-ended motion.

Each clip remains a separate asset with its own recipe and charge. Link shots when the edge describes their story order; recipe references already record provenance.

Review and export

Open each returned web URL, inspect the full clip and star the usable shots. Generate a new asset from a failed or unsuitable card instead of changing its media. Export the space record when the sequence is ready for an external editor or build script.

The browser timeline model is currently unavailable. Export or download the ready clips for the final cut.