Cost to build a 25-video swipe file on Sume: about $4

A swipe file of 25 competitor videos costs about $4 on Sume: $3.75 to import them at $0.15 each, nothing for probes and stills, and about one cent per clip for a transcript. The imports are the only large line, so decide which clips are worth keeping before you pay for them.

This page prices each step of the pipeline from the published rate card, tells you which steps are free, and shows a short script that totals the cost before you run it. The transcript line uses the video inspect rate of $0.01 per audio minute, with a one-minute floor on the reserve.

The $4 figure is a planning number for a first pass. It assumes every import is accepted, every clip is under a minute and nothing is retried. Real runs differ at the edges, and the usage ledger is where you read what actually happened.

What does each step of a swipe-file pipeline cost?

A swipe file needs four things per clip: the file in your workspace, its facts, a few stills, and the words spoken. Only the first and last have a price.

There is also a practical limit on what you can import. Only public TikTok and Instagram video URLs are accepted, and other hosts are rejected, so a swipe file built from YouTube links needs a different intake path.

Why are the imports the expensive part?

Import is a flat $0.15 per accepted import, so its cost does not depend on clip length. A 7-second clip and a 55-second clip both cost 15 cents, while a transcript for either is about a cent.

That makes shortlisting the best saving. If you search the trending catalog first, you get metadata and public watch URLs without importing anything, and you can filter on the numbers you care about. Import only the clips that pass. Cutting a 25-clip list to 15 saves $1.50 per round.

A cheap habit is to write the transcript only for clips you have already looked at. The stills from the free frames route show the opening shot, the on-screen text and the pacing, which is often enough to discard a clip without ever paying for its words.

How long can a clip be before the transcript cost grows?

The transcript is billed per audio minute. The catalog entry shows a one-cent minimum and a ten-cent maximum, which matches the documented 600-second ceiling on the duration hint. A 4-minute reference is therefore a few cents, not dollars.

If a clip has no speech, a transcript is wasted. Probe first, which is free, and look at the audio facts before you ask for words.

Keep the balance check in the same script as the plan. If the wallet cannot cover a reserve, Sume refuses the request with 402 insufficient_credits before work starts, and failing early is cheaper than discovering it at clip 19.

Can I total the cost before running it?

Yes. The arithmetic is simple enough to keep in your own code, and checking the balance first avoids a 402 halfway through the loop.

import os
import requests

IMPORT_USD = 0.15
TRANSCRIPT_USD = 0.01  # per audio minute, 1 minute floor
clips = 25
minutes_each = 1

plan = clips * (IMPORT_USD + TRANSCRIPT_USD * minutes_each)
r = requests.get(
    "https://api.sume.com/v1/balance",
    headers={"Authorization": f"Bearer {os.environ['SUME_API_KEY']}"},
    timeout=30,
)
print(f"plan: ${plan:.2f}")
print(r.status_code, r.json())

What does this not include?

• Rights. Importing a post gives you the file for analysis; whether you may reuse the footage is a separate question.
• Search calls. The trending search that finds the clips has its own per-call price on the pricing page.
• Rework. Imports that you retry without an idempotency key can create a second job, and the second job is a second $0.15.