Which Sume video models accept 30 seconds? A Python filter

Ask GET /v1/videos/models which models list a 30-second duration, instead of trusting a launch post. A short Python script, plus what the docs name.

4 min readSume
All posts

Do not trust a launch post for clip length; ask the catalog. GET /v1/videos/models returns supported_durations for every model, so a six-line Python filter tells you which ones accept 30 seconds today. Sume's docs name four that reach 30 seconds, and each is a different kind of tool.

New models launch weekly, and 30-second clips are now a headline spec across vendors. The catalog is the one place that reflects what your key can call. Facts here are from Sume's video generation docs and Video Router docs, read 2026-10-02.

How do I filter the catalog?

The response has a data array. Each model carries supported_durations (whole seconds), supported_resolutions, supported_aspect_ratios and generate_audio. The filter keeps models whose list contains 30 and prints their range.

import os
import requests

response = requests.get(
    "https://api.sume.com/v1/videos/models",
    headers={"Authorization": f"Bearer {os.environ['SUME_API_KEY']}"},
    timeout=30,
)
response.raise_for_status()

for model in response.json()["data"]:
    durations = model.get("supported_durations") or []
    if 30 in durations:
        print(model["id"], min(durations), "to", max(durations), "s")
        print("  ratios:", model.get("supported_aspect_ratios"))

Which models does the documentation name?

The docs describe limits per model and say to read the catalog rather than assume one envelope. Ids and ranges below come from the docs; a model listed only when its provider is configured may be absent from your catalog response.

30-second models named in Sume docs, read 2026-10-02
Model idRange in docsWhat it does
seedance-2.54 to 30 s at 480p, 720p, 1080pClip generation
wan-3.02 to 30 sClip generation
higgsfield-genjutsu4 to 30 s, 480p or 720pMotion transfer: one source video plus 1 to 8 reference images
h3-max-recast5 to 30 s, 768p or 1080pSwaps people in a source video for 1 to 4 person photos

Is every 30-second model a prompt-to-video model?

No. Genjutsu and H3 Max Recast take a source video, so their length follows the source. Only choose them when you have footage to transform. Every other model in the catalog tops out at 15 seconds, and sume/auto makes 3 to 10 second clips.

What does the filter not tell me?

It does not tell you price, quality or speed. Check the model's pricing_skus in the same response, and make a short test clip before you queue a batch. Sume bills provider list times 1.25, reserved on submit.

Sources

Related posts

More in Developers

All Developers posts

Written by Sume