What does $50 of Sume balance buy? Avatar seconds, music, Timeline

Fifty dollars of Sume balance buys, at the listed rates (read 2026-10-03), about 271 seconds of Standard avatar video, 204 seconds on Plus or 90 seconds on Max, or 400 music generations at $0.125, or 500 minutes of Timeline output at $0.10. In practice you mix them. A good first $50 is one avatar, three 20-second Plus clips, ten music candidates and a Timeline join, with room for retakes.

These are ceilings per category, not a mix. The useful question is what you want to have at the end, so the sections below price three realistic deliverables inside $50.

Single-product ceilings

The rate card at sume.com/pricing/api is the source of truth, and the Sume docs confirm the same units. Avatar creation is a flat $0.95 per avatar and is not included in the seconds below.

Three realistic mixes

The first is a launch teaser: one avatar ($0.95), two 30-second Max clips ($33.00), five music candidates ($0.63) and one Timeline join ($0.10) come to $34.68, leaving $15 for retakes. The second is a content test: one avatar, six 20-second Standard clips ($22.08), six music tracks ($0.75) is $23.78. The third is a multilingual pilot: three avatars ($2.85), six 25-second Plus clips ($36.75), and three music tracks ($0.38) is $39.98.

Each mix keeps at least 20 percent of the balance unspent. That buffer is not waste: failed or retaken jobs are real spend, and a 402 mid-batch costs more in time than the buffer costs in dollars.

Check it in code

This snippet prices the three mixes so you can edit the quantities.

import asyncio

R = {"std": 0.184, "plus": 0.245, "max": 0.55, "music": 0.125, "avatar": 0.95, "tl": 0.10}

async def main() -> None:
    mixes = {
        "teaser": R["avatar"] + 60 * R["max"] + 5 * R["music"] + R["tl"],
        "test": R["avatar"] + 120 * R["std"] + 6 * R["music"],
        "pilot": 3 * R["avatar"] + 150 * R["plus"] + 3 * R["music"],
    }
    for name, cost in mixes.items():
        print(f"{name:7s} ${cost:6.2f}  left ${50 - cost:5.2f}")

asyncio.run(main())

Where the balance lives

Top up on the dashboard's Billing & subscription page, and read the current balance with GET /v1/balance, as described in Billing and credits. Concurrency is set by plan rather than by balance, so $50 will not make jobs run faster, only let you run more of them.

What $50 does not buy

It does not buy more concurrency: the processing limit is set by plan. It does not buy faster jobs on a given tier; Standard is the fastest path by design. And it does not cover paid items I have not listed, such as image generation or video routers, whose prices are metered per job or per model and are in the catalog rather than on a flat rate card. Use GET /v1/catalog or the pricing page for those before you add them to a mix.

If you are testing, a sensible split is to spend about half of the first $50 on the actual deliverable and keep the other half for retakes and experiments. Revisit that ratio after the first week, once you know your real retake rate.