Cost of 12 avatar clips (4 each at 15, 30, 45 s) on Sume

Twelve avatar clips, four each at 15, 30 and 45 seconds, add up to 360 seconds of video. At Sume's listed per-second rates for productless clips that is $66.24 on Standard, $88.20 on Plus and $198.00 on Max, plus a one-time $0.95 to create the avatar (read 2026-10-03). Plus is the default tier when you omit quality, so an unmodified request lands in the middle row.

This is a typical weekly content calendar for one presenter: short hooks, mid-length explainers, and a few longer walkthroughs. The arithmetic is simple, but it is easy to get wrong in a budget meeting because the rate is per second of video, not per clip, and the avatar is paid for once, not per clip.

The rates behind the math

Sume's API pricing page lists Avatar 1.0 talking video per second, split by quality tier and by whether you attach a product image. The avatar video docs confirm the three tiers (standard, plus, max), that plus is the default, and that a script or multi-scene plan must come out at 4 to 60 seconds. Creating the reusable avatar is a separate flat charge per avatar, described in Create new avatar.

The figures below use the no-product rates. If your clips show a product image, the per-second rate is slightly higher on every tier; see what a product image adds.

The 12-clip total by tier

Each row below multiplies the 360 planned seconds by the tier rate and adds the single avatar creation charge. The clip mix (4 + 4 + 4) matters only through the total seconds; twelve 30-second clips would cost the same.

Reproduce it in a few lines

The script below is plain arithmetic, so it runs anywhere. Change the clips list to match your own calendar. Treat the output as a planning figure: the number to trust after a run is the spend in your workspace usage, which you can read through GET /v1/usage as described in Billing and credits.

# Cost of a 12-clip avatar calendar at documented per-second rates (USD).
RATE = {"standard": 0.184, "plus": 0.245, "max": 0.55}  # no product image
CREATE_AVATAR = 0.95
clips = [15] * 4 + [30] * 4 + [45] * 4
seconds = sum(clips)
for tier, rate in RATE.items():
    total = seconds * rate + CREATE_AVATAR
    print(f"{tier:9s} {seconds} s -> ${total:,.2f} incl. avatar creation")

Three ways the bill moves

If you want to see the first frame before paying for the full render, use avatar video previews and then call generate-video only on the approved ones.

• Tier mix: rendering drafts on Standard and only finals on Max is cheaper than Max for everything. With 12 clips, moving 4 of them from Plus to Standard saves $7.32 if they are 30 seconds each.
• Retakes: a rejected clip is a new paid job. Budget a retake allowance on top of the planned seconds rather than hoping every script lands.
• Duration estimate: Sume estimates the video duration from your script, so a longer script costs more even if you intended a shorter clip. Keep each script inside the 4 to 60 second window.

Choosing a tier for each clip

The cheapest plan is rarely one tier for everything. The documentation describes Standard as the fastest Sume execution path, Plus as the balanced default and Max as the highest quality tier with slower turnaround. A reasonable split for the calendar above is to use Standard for the 15-second hooks that you will run several variants of, Plus for the 30-second explainers, and Max only for the longest walkthroughs that carry the most viewing time.

With that mix the 12 clips cost $139.44 instead of $198.00 for Max everywhere. The point of the exercise is not that one tier is correct, it is that the per-second meter lets you spend quality where it matters. If you want to preview composition before choosing, the preview stage lets you approve first frames and then set quality only on the final render, so a preview you approved is not wasted when you change tier.

What to record after the run

Planning figures are only useful if you compare them with what happened. After the batch finishes, read the workspace usage through the documented GET /v1/usage endpoint and the balance through GET /v1/balance, then divide the spend by the seconds you actually rendered. If the gap is more than a few percent, look at which scripts were estimated longer than you meant, because the duration is estimated from the script text.

Keep a one-line record per clip: script length in seconds, tier, whether a product image was attached, and whether it was accepted or retaken. After a month that record is your own rate sheet, and it is worth more than any generic calculator.