Avatar 1.0 quality: standard for drafts, max for the final weekly cut

If you publish a weekly avatar video, treat the quality field on the Avatar 1.0 route as a draft-versus-final switch: standard is the fastest Sume execution path, plus is the default when you omit the field, and max is the highest tier with slower turnaround. Render the script at standard, read it, then send the same request again at max for the version you ship.

The docs state the behaviour of each tier, not a measured time or a quality score, so this post gives you no benchmark. The cost for each tier is on Sume's API pricing page; read it there before you budget a season.

The three values

A draft-then-final loop

Use two idempotency keys that differ only in the quality, so a retry never re-bills a draft and the final is a deliberate second job. Everything else in the body must be identical, or you are comparing different videos.

def body(script, quality):
    return {"avatar_handle": "YOUR_AVATAR_HANDLE", "script": script,
            "aspect_ratio": "9:16", "quality": quality}

def key(series, episode, quality, version=1):
    return f"{series}:e{episode:02d}:{quality}:v{version}"

draft = (key("weekly", 4, "standard"), body("Three things changed.", "standard"))
final = (key("weekly", 4, "max"), body("Three things changed.", "max"))
assert draft[1]["script"] == final[1]["script"]
print(draft[0], final[0])

Cheaper ways to check a frame

Before any full render, an avatar video preview shows first-frame stills. You then call generate-video on the preview id. That is the docs' path for "review first, pay after", and it is separate from the standard tier.

Limits that do not change with quality

• Length stays 4 to 60 seconds for every tier.
• Resolution is currently 720p for every tier.
• Aspect ratios stay 1:1, 3:4, 9:16, 4:3, 16:9; the default is 9:16.
• Sume does not publish a different set of avatars per tier. The avatar comes from avatar_handle, which must be ready.