Filter the Sume video catalog in Python for 1080p and audio references

Call GET /v1/videos/models, read the data array and filter on supported_resolutions and supported_input_references. The script below keeps models that list 1080p and accept audio_url references, and prints each one's id and durations.

This beats a hard-coded table, because the catalog is what validation uses. A value the catalog does not list gets a 400, so the catalog is the one source of truth about what you can send.

What each catalog row carries

Each VideoModel has id, supported_resolutions, supported_aspect_ratios, supported_durations, supported_frame_images, supported_input_references, generate_audio and pricing_skus. supported_input_references is the one Sume adds on top of the OpenRouter shape, because some models reject references that others accept.

pricing_skus maps each SKU to a price string, and the SKU unit differs by model (for example per 1,000 video tokens), so compare like with like.

The script

It uses only the standard library. Change the two filter lines to ask a different question, such as models that accept video_url references or a 15-second duration.

import json, os, urllib.request

req = urllib.request.Request(
    "https://api.sume.com/v1/videos/models",
    headers={"Authorization": "Bearer " + os.environ["SUME_API_KEY"]},
)
with urllib.request.urlopen(req, timeout=30) as r:
    models = json.load(r)["data"]

for m in models:
    refs = m.get("supported_input_references") or []
    if "1080p" in (m.get("supported_resolutions") or []) and "audio_url" in refs:
        durs = m.get("supported_durations") or []
        print(m["id"], min(durs, default=None), max(durs, default=None), m.get("pricing_skus"))

Reading the result

The output has one line for each model that passes both tests. A model that is absent either lacks 1080p or rejects audio references. sume/auto is never in the list, because it is not a model, but the guide documents it.

If a row has a null or empty list for a field, the script treats it as not supported. That is the safe reading, since a value the catalog does not advertise returns a 400.

Use it in a pre-flight check

Run the filter on startup in a service that picks models for users. Cache the result for a few minutes, and pick from it, instead of shipping a table that ages. When a model is added, the filter picks it up with no code change.

Pair it with the pricing field to compare the models that pass, within the same SKU unit.

• Fail early with a clear message when no model passes the filter.
• Log the catalog's model ids with each job so a later review can see what was offered.
• Do not send a field the row does not list, because it will be a 400, not a silent drop.