Which Sume operations cost under 10 cents? The cheap-steps list

Nine Sume operations list at 10 cents or less: audio detach at $0.01, a transcript at $0.01 per audio minute, a timeline audio mix at $0.01, trim, filter and compose at $0.02 each, background removal at $0.0225 per image, text-to-speech at $0.0475 per 1,000 characters and a timeline render at $0.10 per output minute. Generation steps sit well above that, as the figures below show.

These are the glue steps of a video pipeline. They are cheap enough that you can stop counting them in a small project, but at 10,000 clips a month they add up, so it helps to know which ones are flat and which scale with length.

Everything here is a list rate read from the catalog on 2026-10-02.

What does each cheap operation cost?

The table lists the unit Sume bills in, because that unit decides how the price scales. Flat per-job prices do not grow with clip length. Per-minute and per-character prices do.

If you need to budget a loop, flat prices are the easy case: multiply by the number of jobs. For the scaling ones, multiply by minutes or characters and then remember the reserve ceiling, which is what the wallet must be able to cover at submit time.

Which of these are flat and which scale?

Detach, audio mix, trim, filter and compose are flat per job. A 5-second trim and a 4-minute trim both cost two cents on the catalog. Transcripts, text to speech and timeline renders scale with audio minutes, characters or output minutes.

For the scaling ones, the catalog also publishes a reserve ceiling. A transcript reserves between one and ten cents, and a timeline render reserves between ten cents and three dollars. The reserve is what Sume holds at submit; the captured amount follows the real duration.

What does a 25-clip cleanup cost?

Suppose you take 25 imported clips, trim each to a hook, strip the audio, and render one 60-second compilation. That is 25 trims ($0.50), 25 detaches ($0.25) and a 1-minute timeline render ($0.10), for $0.85 before any import or generation.

Add a voiceover for the compilation: a 600-character script is $0.0285 at the TTS rate. The glue stays under a dollar even with the voice.

Because the flat steps cost so little next to generation, it is usually right to trim before you generate or analyse, and not after.

RATES = {
    "trim": 0.02,
    "detach": 0.01,
    "render_min": 0.10,
    "tts_per_1k_chars": 0.0475,
}
clips, render_minutes, chars = 25, 1, 600

total = (
    clips * RATES["trim"]
    + clips * RATES["detach"]
    + render_minutes * RATES["render_min"]
    + chars / 1000 * RATES["tts_per_1k_chars"]
)
print(f"${total:.2f}")

Where does the real money go?

Not here. Generation steps are priced per second, per clip or per image at rates that start in the tens of cents and climb into dollars. Avatar Video, for example, lists $0.184 per second on Standard.

So when you trim a budget, look at the generation line first. The cheap steps are worth keeping honest mainly because they run in loops. Ten thousand two-cent trims is $200, which is a number you notice.

The same logic applies to image work. Removing a background is $0.0225 per image, but generating a new image lists at about $0.22 on the image router estimate, roughly ten times more. Reuse and edit before you regenerate.

How do I confirm these rates are current?

Read the API pricing page, which Sume generates from the same constants its billing path uses. The numbers above were read from that catalog on 2026-10-02, and a price change would show there first.