How many Seedance 2.5 clips fill a 3-minute Reel: six at 30 s

A three-minute Reel is 180 seconds, and seedance-2.5 on the Sume Video Router accepts 4 to 30 seconds per clip, so you need at least six clips at the maximum length and more if you want shorter shots. Sume documents no single call that returns three minutes; you generate the clips, then join them in a Timeline 1.0 render. Instagram's own Reels page says Reels can be up to 20 minutes, but it also says videos longer than 3 minutes may not be recommended to new audiences (read 2026-10-03), which is why 180 seconds is a useful target.

Clip counts at different shot lengths

Shots under 4 seconds are not available on this model. If your edit rhythm needs 2 to 3 second beats, wan-3.0 documents a 2 to 30 second range, or you can generate longer clips and cut them with video trim.

Submit the clips

The script below submits six 30-second 9:16 jobs, each with its own idempotency key, and prints the job ids. Prices vary by resolution, so confirm the live per-second rate in GET /v1/catalog and multiply by 180 before you run it at scale. A 480p pass is a cheaper way to check the story before a final render, though a rerun is a new generation, not a refinement of the earlier one.

Each job is asynchronous: poll GET /v1/jobs/:id/status, then read /result for the clip's hosted media.sume.com URL.

import json, os, urllib.request

API = "https://api.sume.com/v1"

def post(path, body, key):
    req = urllib.request.Request(
        f"{API}{path}",
        data=json.dumps(body).encode(),
        headers={
            "Authorization": f"Bearer {os.environ['SUME_API_KEY']}",
            "Content-Type": "application/json",
            "Idempotency-Key": key,
        },
        method="POST",
    )
    with urllib.request.urlopen(req) as res:
        return json.load(res)

prompts = [
    "Scene {n}: a vertical UGC-style desk setup, natural light, slow handheld push-in",
] * 6

jobs = []
for n, prompt in enumerate(prompts, start=1):
    job = post(
        "/video-router/generate",
        {
            "model": "seedance-2.5",
            "prompt": prompt.format(n=n),
            "resolution": "480p",
            "duration": 30,
            "aspect_ratio": "9:16",
            "mode": "async",
        },
        f"three-minute-reel-scene-{n}",
    )
    jobs.append(job["request_id"])

print(len(jobs) * 30, "seconds requested")
print(*jobs, sep="\n")

Stitch and keep the story together

Joining is a render, not a generation: Timeline 1.0 takes ordered video[] slots with optional transitions of one second or less, bills $0.10 per ceiling output minute, and renders 256 to 2160 pixels per edge at 24, 25, 30 or 60 fps. Run the free /timeline-1.0/plan call first to confirm the layout before you pay for the render.

The harder problem is continuity. Each clip is generated on its own, so a character or product can drift between scenes. Anchor every prompt on the same described subject, keep the camera language consistent, and check the joins in the first frames of each clip. Sume makes no promise that separately generated clips will match; budget a retry on the clips that do not.

Is a three-minute generated Reel a good idea

Keep a ledger as you go: scene number, prompt, idempotency key and job id in one file. When a clip fails review you will want to resubmit just that scene with a new key and a changed prompt, and the ledger tells you which jobs the final Timeline should reference.

The length is a choice, not a goal. Instagram's page lists 20 minutes as the ceiling and describes a 3-minute line past which new audiences may not be shown a Reel (read 2026-10-03), so a 90-second Reel stays well inside it and needs only three 30-second clips or about eight 12-second ones.

Generated footage also costs more to get wrong the longer it runs. A mistake in clip five means a retry of one clip, not of the whole film, which is the real benefit of building the Reel from separate jobs: you pay to regenerate one scene, and the idempotency key for the others stays the same, so a re-run of your script does not create duplicates.

Start with a storyboard of the shots you would film if you had a camera, give each a number, and send the script once for a single scene. Look at that scene before you submit the other five.