How much to top up a Sume wallet: what $10 and $1,000 buy
Sume wallet purchases run $10 to $1,000. What each buys in avatar seconds, music, cutouts and captions, and a sizing rule: burn plus the largest hold.

Size a Sume top-up as your expected spend until the next top-up plus the largest single hold you will place, then round up inside the $10 to $1,000 range that the pricing FAQ gives for on-demand purchases. A $10 purchase buys about 40 seconds of Avatar Plus video; $1,000 buys about 4,081 seconds.
Top-ups are made in the dashboard. The Billing and credits page says the public API exposes balance and usage reads, not a top-up endpoint, so a person has to be in the loop.
What does a purchase buy?
Divide the purchase by the published rates from the API pricing page. Whole units only; the rates are read 2026-10-03.
| Item | Rate | $10 buys | $1,000 buys |
|---|---|---|---|
| Avatar Plus video, no product | $0.245 per second | 40 seconds | 4,081 seconds |
| Music track | $0.125 each | 80 tracks | 8,000 tracks |
| Background removal | $0.0225 per image | 444 images | 44,444 images |
| Image upscale | $0.20 per image | 50 images | 5,000 images |
| Caption job (up to 60 s) | $0.20 per job | 50 jobs | 5,000 jobs |
| Text to speech | $0.0475 per 1,000 characters | 210,526 characters | 21,052,631 characters |
| Speech transcription | $0.01 per minute | 1,000 minutes | 100,000 minutes |
Why add the largest hold?
A submit reserves its estimate first. The admission docs say a request fails with 402 insufficient_credits before provider work starts when Sume cannot reserve that cost. So a wallet of $20 with a spend plan of $15 can still refuse a single job whose reserve is higher than the $20 on hand, such as a long avatar clip: 120 seconds of Avatar Plus reserves $29.40.
Every job in a batch holds its estimate until it settles, so a wave of 24 queued jobs holds 24 estimates at once. Work out the wave, not the average. Largest hold one Sume call can place lists the per-endpoint ceilings.
What is a simple sizing rule?
Take the burn per day from your own snapshots (see forecasting days of balance left), multiply by the days until someone can next top up, and add the largest hold in your next batch. A team that spends $25 a day, can top up every Monday and runs one batch with a $30 maximum hold needs about 7 x $25 + $30 = $205 on hand, so a $250 purchase fits; a $1,000 purchase is idle money for most of a month.
What about budgeting in another currency?
The balance response carries a currency field, and the source defaults it to USD; the docs call the balance USD-denominated. The pages cited here do not describe conversion, so treat the card issuer's rate on the day of purchase as the rate and keep a margin in your budget for movement. Do not budget from a rate you looked up earlier in the month.
What should you not rely on?
Do not assume the purchase range is permanent; it comes from the public FAQ as of the read date. If you need more than $1,000 at a time, or invoice terms, the Enterprise route on the pricing page is the path: it lists custom volume pricing and invoice or PO billing, with a contact address rather than a price.
Sources
Related posts
More in Pricing
- Ideogram 4 self-hosting: $300 a month licence vs per-image API
Ideogram's self-serve licence is $300 a month for up to 10,000 images, before GPU time. Where it beats a per-image API such as Sume's, and where it does not.
- Is ElevenLabs v4 Turbo half the price of v4? Rates and monthly cost
Yes: ElevenLabs lists v4 Turbo at $0.011 vs v4 at $0.022 per 1K characters until Oct 12, then $0.04 vs $0.08. A 1M-character month is $11 vs $22.
- Kling 3.0 Motion Control price per second and a 30-second clip
fal lists Kling 3.0 Standard Motion Control at $0.126 a second. On Sume it is $0.1575 a second, rounded up per second: $4.725 for the 30-second maximum.
- Luma Ray 3.2 360p draft costs $0.06 for 5 seconds: the draft ladder
Luma lists a 360p draft tier for Ray 3.2 at $0.06 per 5 seconds. The full ladder to 1080p, drafts per final, and the cheapest Sume draft route.
Written by Sume