Compose mixed media after each still and take is allowed to freeze

compose_video_with_overlay is the merge that merge_movie_scenes is not: images and videos, in listed order, with optional per-clip trims. Each base_media item is an image (needs duration_sec) or a video (duration_sec or start_sec/end_sec, not both). It is a fixed-order concatenation, not a multi-track NLE. That is a finishing assembly of assets you are keeping. A hold of a rejected still for three seconds is three seconds of a board you already threw away.

Combine photos and clips into one video is the job. Billing scales with items and output length. estimate_cost on a long timeline. If the base is a single video with no stills, you wanted merge or a trim.

Every item is a lock

The stills have to be keepers. The clips have to be keepers. The hold durations have to match a clock you believe. People compose first to "see the sequence," then replace still three, then re-trim clip two, then compose again, then caption the second compose, then replace still three again. Overlay type and music on the child, once.

The slideshow maker is often where the stills came from. Do not compose a reel of a carousel you still intend to replace_slideshow_slide. Convert the slideshow, or export the stills you have signed, then compose if you also have motion.

Order

1. Sign each still and each clip.
2. List them in order with holds and trims you meant.
3. Compose once.
4. Score and caption the result.

Do not mix duration_sec and a start/end range on the same video item. The tool treats those as mutually exclusive.

The AI video generator still has to produce clips worth listing.

FAQ

Can I use this as a storyboard by dropping in temp stills?

You can build a paper edit in a doc. A billed compose of temps trains you to keep the temps. Sign, then compose.

Should music be generated before the compose?

After you know the output length. A 40-second compose and a 12-second bed is a mismatch you created by scoring early.

Is picture-in-picture the same tool?

PiP on a single base is create_ugc_video_overlay. Mixed concatenation is this. Both wait for keepers. If you need both, you are in a heavier EDL — still later.

Why not edit_video for everything?

You can, for a full assembly with captions and mix in one call — on keepers, with preview:true first. This page is the mixed-media concatenate. Pick the smaller tool when it fits. Neither is a first step.