Upscale a 45-second clip: 30 s maximum, so two trims, $0.445
Sume Video Upscale takes at most 30 s at $0.009 a second. A 45-second clip needs two trims ($0.04) and two upscales ($0.405): $0.445, or $0.545 with a render.

Sume Video Upscale 1.0 is $0.009 per video second with a maximum of 30 seconds, so a 45-second clip cannot go in as one job. Cut it into a 30-second part and a 15-second part with two video trim jobs, upscale each, and the bill is $0.04 + $0.27 + $0.135 = $0.445. Joining the two results back on Timeline 1.0 adds one render at $0.10 for a total of $0.545.
The line items
Each step has a public flat or per-second rate. Trim is $0.02 per job and the timeline render is $0.10 per output minute, rounded up to a whole minute with a one-minute minimum.
| Step | Rate | Quantity | Cost |
|---|---|---|---|
| Video trim, 0 to 30 s | $0.02 per job | 1 | $0.02 |
| Video trim, 30 to 45 s | $0.02 per job | 1 | $0.02 |
| Video upscale, 30 s part | $0.009 per second | 30 s | $0.27 |
| Video upscale, 15 s part | $0.009 per second | 15 s | $0.135 |
| Timeline render, 45 s output | $0.10 per minute, rounded up | 1 minute | $0.10 |
| Total with the render | $0.545 |
Why duration_seconds matters
If you do not send duration_seconds, Sume reserves 5 seconds, which is $0.045, then captures the actual cost. Sending it makes the hold match the job. The accepted range is 1 to 30 seconds, and the request also takes an upscale_factor of 1.1 to 4 and an enhancement_tier of fast, standard or pro.
- The 30-second cap is on the input, so a 31-second clip needs two jobs.
- Use exact precision on the trims so the parts meet at 30 seconds.
- The price is the same for every tier named in the request schema.
The first part
Trim the first 30 seconds, wait for the job, then pass its video_url to the upscale call.
curl -X POST https://api.sume.com/v1/video-trim \
-H "Authorization: Bearer $SUME_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: part-a-trim" \
-d '{
"video_url": "https://media.sume.com/artifacts/artf_demo/clip45.mp4",
"start": 0,
"duration": 30,
"precision": "exact"
}'Check before you scale up
A 30-second part at $0.27 is the most one job can cost. Try a 5-second part first at $0.045 to see the result, then run the full parts. The catalog gives a maximum job estimate of 27 cents before the Sume agent fee, which matches 30 x $0.009.
Scaling it up
For 20 such clips the bill is 20 x $0.445 = $8.90 without a render, and 20 x $0.545 = $10.90 with one render per clip. If the 45-second clip is cut at 22.5 seconds instead, both halves fit the limit and cost 45 x $0.009 = $0.405 in upscale either way, because the price is per second, not per job. Only the trim count and the reserve differ. Splitting evenly keeps each hold at $0.2025, rather than a $0.27 hold on one part.
Sources
Related posts
More in Media tools
- $5 of lip sync: 50 s on H3 Max 768p, 26 s on Fabric 720p
A $5.00 balance buys 50 audio seconds of H3 Max Lip Sync at 768p ($0.10), 80 at 480p, or 26 whole seconds of VEED Fabric at 720p ($0.1875) on Sume.
- A 600-character script to a talking still: $0.03 TTS plus $3.80 Fabric
600 characters of TTS cost $0.0285 on Sume; if the audio measures 38 s, VEED Fabric adds 38 x $0.10 = $3.80 at 480p or $7.125 at 720p. Total $3.83 or $7.15.
- 9-second Kling motion control clip: $1.42, or $1.64 finished
Kling 3.0 Motion Control on Sume bills $0.1575 per output second: 9 seconds is $1.4175. With Music 1.0 and a 1-minute render it is $1.6425.
- One 95-second LinkedIn video, two trims: 90 s in-stream, 30 s feed
LinkedIn lists in-stream ads up to 90 seconds and recommends 15 to 30 seconds. From a 95-second source, trim start 0 for 90 s and start 20 for 30 s: $0.04.
Written by Sume