A 3-minute Short from twelve MiniMax H3 clips: $13.80 at 768p

A 180-second vertical Short built from twelve 15-second MiniMax H3 clips at 768p costs $13.50 in generation ($0.075 per second x 180) plus $0.30 for three Timeline render minutes: $13.80. At 480p the same twelve clips cost $11.25.

Shorts can run up to 3 minutes (YouTube Help, read 2026-10-09), so 180 seconds is the longest legal Short. H3 is the cheapest per-second vertical model in this lane, but its 15-second ceiling is what drives the clip count.

Cost by resolution

MiniMax H3 (not Max) has native 480p and 768p; 768p is its first-class resolution and is not 720p. It accepts 5 to 15 seconds and 9:16.

Why twelve slots is the edge of a single render

Timeline auto-chunks a render past 12 segments, and it refuses strategy single above 12 slots (render_strategy_unsafe). Twelve clips sit exactly on the line, so the default render strategy works without any override. A thirteenth slot, for example a short outro, makes the render chunked.

Declared starts are authoritative: the 12th clip starts at 165 and runs 15 seconds to land on 180. Set audio.duration_seconds to 180; coverage can stop at most 0.5 seconds before the end of the spine.

# twelve slots, 15 s each, starts 0,15,...,165
python3 - <<'PY'
import json
slots=[{"source_url":f"https://media.sume.com/artifacts/artf_demo/clip{i+1}.mp4","start":i*15,"duration":15} for i in range(12)]
print(json.dumps({"audio":{"mode":"silence","duration_seconds":180},"video":slots,"output":{"width":1080,"height":1920}}))
PY

What to check before you run twelve jobs

• Run one clip first. Twelve clips at $1.125 each is a $13.50 mistake if the prompt style drifts.
• Each job must be a whole second from 5 to 15; a 180-second target divides evenly, but a 175-second one does not.
• Plan first: POST /v1/timeline-1.0/plan returns billable_minutes and estimated_cost_usd_micros without creating a job.
• Pass a reference image to every request if the same character must appear in all twelve.