Which Sume media steps have a flat price and which are metered
Trim, filter, detach, timeline and captions have fixed public prices. Frames and inspect are billed by Modal compute. Use this list before you budget.
Most Sume media steps have a flat public price: trim $0.02, filter $0.02, audio detach $0.01, timeline audio $0.01, compose $0.02, captions $0.20, and timeline render $0.10 per started minute. Video frames and video inspect are different. Sume bills them by their own Modal compute, so budget them as metered and read the receipt.
The flat-rate list
The trim, filter, detach, audio and render jobs run only worker ffmpeg. The docs say there is no provider inference, which is why the rate can be fixed.
| Surface | Rate | What it does |
|---|---|---|
| Video trim | $0.02 per job | Cut [start, end) from one clip |
| Video filter | $0.02 per encode job; the check is free | Dim, crop or an allowlisted filtergraph |
| Audio detach | $0.01 per job | Audio track to wav or mp3 |
| Timeline audio | $0.01 flat per job | Concat or split audio |
| Timeline compose | $0.02 | Two sources in one frame |
| Timeline render | $0.10 per ceil(output minute) | Assemble clips on an audio spine |
| Video captions | $0.20 per accepted job, up to 60 s | Burn captions on a video |
The metered pair
POST /v1/video-frames reserves its Modal compute ceiling at submit and bills the job at container seconds times the Modal list price times 1.25, plus the platform fee. The bill is never more than the hold. POST /v1/video-inspect works the same way for its probe and stills.
Inspect has one more line. If you set transcribe: true, Sume runs speech-to-text at $0.01 per audio minute (STT_PUBLIC_PRICING). Without a duration_seconds hint it reserves one minute, and the maximum hint is 600 seconds.
Free checks that save the paid call
Two endpoints exist to catch mistakes before they bill. POST /v1/video-filter/check runs the same schema, allowlist and source checks as the encode and returns diagnostics. It creates no job and reserves no credits. POST /v1/timeline-1.0/plan compiles a timeline and returns billable_minutes and estimated_cost_usd_micros without downloading media.
A plan cannot predict short-source pad or loop warnings, and a check that passes can still fail on the box with a structured job error. Treat both as a filter, not a guarantee.
Reading the live price
The numbers in this post come from the docs on the day it was written. Every page points at GET /v1/catalog for the live rate. Use it in code instead of hard-coding a number.
curl -sS "https://api.sume.com/v1/catalog" \
-H "Authorization: Bearer $SUME_API_KEY"Related posts
More in Pricing
- Sume org workspaces run 10 generations at once; Free runs 1
Concurrency on Sume is plan-based: Free 1, Pro 4, Startup 8, Scale 20, with a floor of 10 for org workspaces. Prepaid top-ups do not raise it.
- Sume plan fee plus usage: the monthly bill for 20 to 300 videos
Pro $40, Startup $120, Scale $400, usage billed per call on top. A Python estimate for 20, 100 and 300 short videos, and the plan-fee share.
- Sume Pro $40 or Startup $120 for a two-person team?
Pro is $40 with 4 concurrent jobs; Startup is $120 with 8 and a shared team workspace. How to decide with queue math instead of guesswork.
- Sume Pro vs Startup vs Scale for a 100-clip batch: rounds and queue
For 100 equal clips, usage costs $188.00 on any plan. Plan fee only changes rounds (25, 13 or 5) and how many submit waves fit the accepted-job limit.
Written by Sume