Make segment clips, waiting loops and ending shots (costs credits per second)

Each target is one video clip, made exactly as POST /v1/videos makes one, from the storyboard's own keyframe pictures, model, length, resolution, draft mode and dialogue: a segment (start frame = its first keyframe; end frame = the next keyframe when the join is connected) a decision's waiting loop (starts and ends on the decision keyframe), or a keyframe's ending shot (the last moment of a path, set with updatePanel patch.ending: { prompt, durationS }: starts on that keyframe, no end frame, and plays instead of its still hold). Every prompt for a model with sound asks for natural ambience and dialogue only, no background music. Costs credits per second of video. Quote first with dry_run: true (every target's price and the total, nothing submitted), get the user's OK, then send the same body without it. The total is checked against your balance and the token's daily cap before anything starts; credits are deducted as each clip completes. Each new take is recorded on the storyboard (the first take of a segment becomes its active one; activate: true makes every new one active) and answers its task_id for GET /v1/tasks/{id}. Targets that cannot be made yet (a keyframe with no picture) come back in refused with the reason. A storyboard whose owner approves agent spending answers 403 except for the quote.