2026-10-08

Is Vidu Q4 Preview on Sume? Check the live video catalog in Python

Sume's documented video catalog lists no Vidu model. A short Python check against GET /v1/videos/models confirms it before you build, plus the closest rows.

4 min readSume

All posts

No. The video catalog documented in the Sume repo lists these ids: seedance-2.5, seedance-2-mini, seedance-2, seedance-2-fast, kling-3, wan-3.0, grok-imagine-video-1.5, minimax-h3, minimax-h3-max, gemini-omni-flash-1.1 and h3-max-recast, with higgsfield-genjutsu added only when its provider is configured. There is no Vidu id, and Vidu Q4 Preview launched on 2026-10-07 on Vidu's own web product and API, not through Sume. Since the live list can change, the safe way to be sure is to ask the endpoint.

Check it against the live endpoint

GET /v1/videos/models returns the catalog for your key. The script below fetches it, prints every id, and says whether a Vidu id is present. It exits with a non-zero status if the request fails, so you can run it in CI.

import os
import sys

import requests

KEY = os.environ["SUME_API_KEY"]
r = requests.get(
    "https://api.sume.com/v1/videos/models",
    headers={"Authorization": f"Bearer {KEY}"},
    timeout=30,
)
if r.status_code != 200:
    sys.exit(f"catalog request failed: {r.status_code}")
ids = sorted(m["id"] for m in r.json()["data"])
print(len(ids), "models:", ", ".join(ids))
vidu = [i for i in ids if "vidu" in i.lower()]
print("vidu rows:", vidu or "none")

What the answer means

If the script prints none, then asking POST /v1/videos for a Vidu model returns 404 model_not_found, because the model is neither a catalog id, an alias nor sume/auto. Do not guess an id: the generate route rejects anything that is not in the list.

Closest Sume rows to the Vidu Q4 Preview feature set, read 2026-10-08
You want	Closest Sume row	Documented limit
Many reference images	wan-3.0	up to 10 images, 2 to 30 s, 480p to 1080p
Voice or audio references	wan-3.0, minimax-h3-max	audio references accepted on both
4K output	gemini-omni-flash-1.1	3 to 10 s, 16:9 or 9:16
Clips over 10 seconds	wan-3.0, seedance-2.5	up to 30 s
Auto choice	sume/auto	Omni 1.1 envelope, 3 to 10 s

What to do next

Steps:

Run the check above with your own key and keep the output with your build notes.
Match your brief to the table: references, resolution and length decide the row.
If you want Sume to choose, send model: "sume/auto"; it validates against the Omni 1.1 envelope and fails closed with 400 unsupported_capability rather than routing elsewhere.
If you need Vidu itself, use Vidu's API at the endpoints on its documentation.

Why a live check beats a list in a blog post

The list in this article is a snapshot of the repo docs on 2026-10-08. Catalogs move: Sume adds rows, a provider can be gated by configuration, and the Genjutsu row is shown only when its provider is configured for the workspace. A short script that reads GET /v1/videos/models and writes the ids to a file gives you a record you can diff next week.

Run it as a scheduled check if a missing model blocks a launch. If a Vidu id ever appears, the response will also show its supported_resolutions, supported_durations and supported_input_references, which tell you whether the 15-image and 3-audio limits carried over. Until then, build against the rows that are listed.

Why the answer can change

The catalog is code, not a marketing page, so a new row appears only when a provider entry, a builder and a price constant exist for it. That is also why the safest check is the live list rather than a blog post, including this one. Re-run the request on the day you plan a project, and keep the output next to your project notes so a teammate can see what was available when you chose a model.

What Sume does not do

Sume does not proxy the Vidu API and does not alias a Vidu name to another model. Asking for Vidu and getting a different model would break the contract that job.model echoes what you requested.