Sume avatar quality defaults to plus when omitted: what it costs

If you leave quality out of an Avatar 1.0 request, Sume renders on Plus, which costs $0.245 per second without a product image, against $0.184 for Standard and $0.55 for Max (read 2026-10-03). The default is documented on both the avatar video and the preview pages, and it is easy to overlook in code that never sets the field.

Plus is {round((0.245/0.184-1)*100)} percent above Standard, so a batch that you meant for Standard but did not annotate runs {round((0.245/0.184-1)*100)} percent over plan. For 100 clips of 20 seconds that is {m(2000*(0.245-0.184))}.

Where the default comes from

The avatar video docs list quality: "standard" | "plus" | "max" and describe Plus as the default 'balanced quality path', Standard as the fastest Sume execution path, and Max as the highest quality tier with slower turnaround. The preview docs repeat that quality defaults to plus on preview create, and add a useful detail: the final render tier can be overridden at generate-video, while preview stills are tier-independent.

So the default does not trap you in the preview flow, as long as you remember to override on the final call when you want a cheaper or better tier.

Make the choice explicit

The remedy is a one-line change: always send quality from a named constant in your code, and fail a code review that adds an avatar call without it. If the tier varies by use, store it in the same config as the template name, for example draft: standard, publish: plus, flagship: max.

In multi-tenant products, expose the tier in your own UI rather than hiding it. Customers who see 'Draft' and 'Final' learn to use the cheaper path for review.

import asyncio

TIERS = {"draft": "standard", "publish": "plus", "flagship": "max"}
RATE = {"standard": 0.184, "plus": 0.245, "max": 0.55}

def payload(script: str, purpose: str) -> dict:
    # Explicit quality: never rely on the Plus default.
    return {"avatar_handle": "product_host", "script": script, "quality": TIERS[purpose]}

async def main() -> None:
    for purpose in TIERS:
        p = payload("Short test script.", purpose)
        print(purpose, p["quality"], f"${20 * RATE[p['quality']]:.2f} for 20 s")

asyncio.run(main())

Auditing a past batch

If you suspect the default has been costing more than expected, read your usage ledger through GET /v1/usage and divide each avatar spend by the clip seconds. A rate near $0.245 per second on clips you assumed were Standard confirms it. Fixing the code is free, and the savings begin immediately.

Do not downgrade blindly. Compare a Standard and a Plus render of one script first; some content looks the same, and some does not.

The same default in other places

The default appears wherever the avatar video body is accepted, including the preview create call, so a pipeline that creates previews and then final videos should set quality on both or accept that only the final override matters. Remember the docs say preview stills are tier-independent: setting a different tier at preview create does not change the stills, it only becomes the default for the later render.

That means one way to stay safe is to set the tier once, at generate-video, and treat the preview as tier-neutral. Another is to set it in both places from the same constant. What you should not do is leave it to the default and assume it is the cheapest.

Product images change the rate too

If you attach a product_image, the Plus rate is $0.258 per second instead of $0.245, and Standard rises to $0.194 from $0.184 (read 2026-10-03). The difference is small per second, but over a long batch it adds up. Make sure your budget uses the product-image column for clips that include one.