Source sizes that reach 3840x2160 within Sume upscale's 4x limit
Sume video upscale accepts factors 1.1 to 4. 960x540 needs 4, 1280x720 needs 3, 1920x1080 needs 2; 854x480 tops out at 3416x1920. Same $0.009 a second.

To reach 3840x2160 with Sume Video Upscale 1.0, the source must be at least 960x540, because the factor is capped at 4. A 1280x720 clip needs factor 3, a 1920x1080 clip needs factor 2, and a 854x480 clip cannot get there: at factor 4 it ends at 3416x1920, short of UHD width.
The factor range comes from the API reference: upscale_factor and scale_ratio accept 1.1 to 4, default 2. The table below assumes the ratio multiplies width and height, which is how the field is described; confirm the delivered size on your own output.
Factor needed for 3840x2160
Divide 3840 by the source width. If the answer is 4 or less, one upscale job is enough.
| Source | 3840 / width | Within 1.1-4? | Result |
|---|---|---|---|
| 854x480 | 4.50 | No | 3416x1920 at factor 4 |
| 960x540 | 4.00 | Yes (the limit) | 3840x2160 at factor 4 |
| 1280x720 | 3.00 | Yes | 3840x2160 at factor 3 |
| 1920x1080 | 2.00 | Yes (the default) | 3840x2160 at factor 2 |
| 2560x1440 | 1.50 | Yes | 3840x2160 at factor 1.5 |
| 3584x2016 | 1.07 | No, below 1.1 | Use another route; factor must be 1.1 or more |
Why the factor does not change the bill
The catalog lists video upscale at $0.009 per video second with a 30-second input maximum, so a 12-second clip costs 12 x $0.009 = $0.108 whether the factor is 1.5 or 4. Choose the smallest factor that clears your delivery size; a bigger factor does not cost more but asks the model to invent more detail.
Send duration_seconds so the reservation matches the clip. Without it the admit step reserves 5 seconds, and the maximum value is 30.
When the source is too small
For a 854x480 source you have two honest options. Accept 3416x1920 as a 1920-line master, or run a first pass at factor 2 (1708x960) and decide whether a second pass is worth the extra $0.009 per second. Sume does not document a quality guarantee for chained passes, so view the result before you commit a batch.
The same job is available to agents as video_upscale_create, listed on the MCP tools and gates page.
Checking the result
After the job completes, run Video Inspect on the output with two or three stills to confirm the size and look at faces and text edges. Inspect bills by its own compute and has no fixed price, so treat it as a small extra, not a free step.
If you deliver to a platform with a minimum size, compare against its own spec page. For example, TikTok's ad spec asks for at least 540x960 on 9:16, which every source in the table clears before any upscale.
- Factor 4 on 960x540 doubles twice: 960 x 4 = 3840 wide and 540 x 4 = 2160 tall.
- Factor 3 on 1280x720: 3840 x 2160, the same target at a lower factor.
- A factor below 1.1 is outside the documented range, so a nearly-4K source should skip the upscale step.
Sources
Related posts
More in Media tools
- TikTok 516 kbps floor: 0.033 bits per pixel at 540x960, 0.008 at 1080p
At 30 fps TikTok's 516 kbps floor is 0.033 bits per pixel at 540x960 but 0.008 at 1080x1920. Why a bigger frame needs a higher real bitrate to look the same.
- TikTok ad bitrate window: 0.516 to 6.67 Mbps for 10 minutes
TikTok wants at least 516 kbps and at most 500 MB. For a 10-minute ad that is a 38.7 MB floor and a 6.67 Mbps average ceiling. The math and what Sume controls.
- TikTok ad captions are white and fixed; burned-in captions differ
TikTok's non-Spark page says ad captions show in white with a font you cannot change. Sume's burned-in caption styles live inside the video, a different thing.
- One 16:9 size for TikTok and LinkedIn ads: 960x540 to 1920x1080
TikTok's 16:9 minimum is 960x540 and LinkedIn's maximum is 1920x1080. 1280x720 and 1920x1080 clear both; 854x480 fails TikTok, 2048x1152 fails LinkedIn.
Written by Sume