Cost of 10 eight-second avatar ad hooks: Standard, Plus, Max

Ten 8-second hooks from a ready avatar cost $14.72 on the Standard tier, $19.60 on Plus and $44.00 on Max, using Sume's per-second rates of $0.184, $0.245 and $0.55 with no product image. If you leave quality out, Sume uses Plus, so the unlabeled request is the $19.60 one.

An ad test is a good fit for per-second billing because every hook is short. This page prices the whole set before you submit it, shows what a product image adds, and explains what to do when the wallet cannot cover the batch. All rates come from the catalog Sume generates its pricing page from, and the arithmetic assumes each clip bills exactly 8 seconds.

Everything here is list pricing from Sume's catalog. The wallet is denominated in US dollars, so you can compare the totals directly with your ad-test budget without converting from credits.

What do 10 eight-second hooks cost on each tier?

Multiply the tier rate by 8 seconds, then by 10 clips. Avatar Video accepts 4 to 60 seconds per request, so 8 seconds is well inside the window.

Does a product image change the price?

Yes. When the request includes a product image the per-second rate rises: $0.194 on Standard, $0.258 on Plus and $0.58 on Max. For ten 8-second hooks that is $15.52, $20.64 and $46.40.

Mix both in one test if only some hooks show the product. Price each group at its own rate, then add the two. The single avatar creation fee of $0.95 is paid once per avatar, not per hook, so it joins the total only if you create a new avatar for the test.

For a small team, the practical split is a cheap exploration round followed by a narrow production round. Spend roughly two thirds of the budget on volume of ideas, and keep the remainder for the clips that earned a second look. That keeps the expensive tier attached to proven scripts rather than guesses.

Which tier should an ad test use?

Sume's docs describe Standard as the fastest execution path, Plus as the balanced default and Max as the highest quality with slower turnaround. They do not publish a quality score, so do not assume a tier difference will move your click rate.

A defensible approach is to test the hook, the script, on the cheapest tier and then re-render only the winner on a higher one. A 10-hook round on Standard costs $14.72, and one Max re-render of the winning 8-second clip adds $4.40. That is $19.12 in total, less than ten hooks on Plus.

The reserve is an estimate, not a charge. If a job fails, the hold is released and the wallet is not debited for it, which is why a failed hook does not eat into the ten-hook total. The refunded row stays in the ledger as a record.

How do I check the bill before I submit?

Sume reserves the estimate when you submit and then captures the real amount or refunds the hold. If the wallet cannot cover the reserve, the request is refused with 402 insufficient_credits before any rendering starts. Read GET /v1/balance first and compare it with your own total.

After the batch, add job_id or run_id to GET /v1/usage and read the summary.debited_usd field. Do not sum the ledger rows yourself, because a refunded row keeps its hold amount.

Because the default is Plus, every request that omits quality is priced at $0.245 per second. If you meant Standard, say so explicitly in the request body. It is a one-field change that cuts the batch by about a quarter, from $19.60 to $14.72.

import os
import requests

RATES = {"standard": 0.184, "plus": 0.245, "max": 0.55}
hooks, seconds = 10, 8
quality = "plus"  # Sume's default when omitted

need = round(RATES[quality] * seconds * hooks, 2)
bal = requests.get(
    "https://api.sume.com/v1/balance",
    headers={"Authorization": f"Bearer {os.environ['SUME_API_KEY']}"},
    timeout=30,
)
print("estimate USD", need)
print(bal.status_code, bal.json())

What should I leave out of the estimate?

• Retakes. A hook you regenerate is a new job at the full rate, so budget 1.3 times the base if you expect to redo some.
• Captions. Inline captions on an avatar video do not create a separate billed caption job, so they do not add a line to this table.
• Face swap. That is a different SKU with its own 15-second ceiling, so do not price it from these rates.