AI animated short film pipeline: shots, Timeline, and a score

Build a short animated film on Sume in three steps: generate each shot with POST /v1/videos, assemble the cut and the audio spine with Timeline 1.0, and make the score with Music Router. Higgsfield's blog lists a September 26, 2026 guide, "How to Make an AI Animated Short Film in 2026" (10 min read), and an October 1 guide on an AI music video for a real artist.

Step 1: shots

Each shot is one job. Pick a model whose limits match the shot: seedance-2.5 accepts 4 to 30 seconds, wan-3.0 2 to 30 seconds, and minimax-h3 5 to 15. Read each model's limits from GET /v1/videos/models. Keep the prompt, the first frame and the aspect ratio the same across shots for continuity. Pass frame_images with a first_frame to start a shot from a still you approved, and store each job id so a crash does not lose work.

Step 2: spine and cut

Timeline 1.0 takes one audio spine and ordered video[] slots and returns one MP4. Run POST /v1/timeline-1.0/plan first. It is unbilled and returns the duration and an estimated cost. Slots can use a transition such as fade or dissolve, up to 1 second.

Render at $0.10 per output minute, rounded up, according to the docs; confirm the rate in GET /v1/catalog. Past 12 segments the render chunks automatically.

Step 3: the score

Music Router takes a prompt of 1 to 5000 characters. Leave model out for sume/music-auto, which uses Lyria 3.5 today, or pin lyria-3.5 or lyria-3-pro. duration fields are rejected, so say the length in the prompt, for example a 2-minute track, or write timed sections in brackets. Put exclusions in the positive prompt.

Lay the track under the dialogue with Timeline's soundtrack bed: gain_db, loop, fade_out_seconds and duck_db for ducking under a real spine.

Keep it honest

Review every shot before assembly, because a bad shot costs a render. Nothing here replaces a story, voices or a writer. It gets you from a script to a watchable first cut, and you decide whether to regenerate.