Render small and upscale, or render large? The credit maths

Every video model in the LUVI catalog charges more for a higher resolution, usually by multiplying a per-second rate. That leaves an obvious question: is it cheaper to render the clip small and enlarge it afterwards? On September 22, 2026 we quoted the Video Upscaler against a real ten-second clip and got 360 credits to 1080p and 480 credits to 2K—36 and 48 credits per second of source video. Here is what that does to the arithmetic, and where it stops being a good idea.

What does upscaling actually cost?

Thirty-six credits per second of the source clip to reach 1080p, and forty-eight per second to reach the 2K tier at 2560×1440. It takes one video, it takes no prompt at all, and the only control is target_resolution with those two values; 4K is not exposed.

That per-second shape is worth stating plainly, because a single figure of 36 credits is easy to read as the price of a clip. It is the price of one second. A ten-second clip came to 360, and a five-second clip works out at 180.

Does rendering small and upscaling really come out cheaper?

Yes, and by a wide margin at the resolutions most people actually use. Take a five-second Wan 3.0 clip, quoted the same day:

• Straight to 1080p: 1,600 credits.
• 480p, then upscaled to 1080p: 400 credits for the render plus 180 for the upscale—580 in total.

The same shape holds on PixVerse v6: 1,150 credits for five seconds straight to 1080p, against 350 at 360p plus 180 to upscale, which is 530. The two-step route comes to about a third of the one-step price on Wan 3.0 and a little under half on PixVerse v6, because the upscaler's per-second rate is small next to a generator's top resolution tier.

There is a second effect that matters more than the credits when you are still deciding what to make. A draft at the bottom resolution is cheap enough to run several times, and you only pay the upscale on the take you keep.

When is this the wrong idea?

When the detail has to be real. An upscaler adds pixels; it does not add information the model never generated. Fine texture, small faces in a wide frame, legible signage, thin lines—if any of those carry the shot, a native render at the higher resolution has something the enlarged draft does not.

Two more limits are worth knowing before you plan around this:

• The ceiling is 2K. The upscaler's contract offers 1080p and the 2560×1440 tier and nothing above it, so a genuine 4K deliverable still has to come from a model that renders 4K.
• Long clips flip the arithmetic. The upscale is per second, so a thirty-second clip costs 1,080 credits to lift to 1080p. Compare that against the generator's own per-second premium for the same length before assuming the two-step route wins.

How does this fit with the estimate?

The credit figure above the Generate button is the ceiling for that run, and the charge settles at the provider's real bill—so the sums here are worst cases on both routes. One practical note: quote the upscale with the video attached. A quote made without the clip cannot see how long it is, and it comes back as the one-second rate, which looks like a bargain and is not a price.

How we checked

• Models: Video Upscaler, with Wan 3.0 Text-to-Video and PixVerse v6 Text-to-Video quoted for comparison.
• Settings: a real 1280×720 clip of ten seconds, quoted to 1080p and to 2K; five-second video renders at 360p, 480p and 1080p.
• Date and what was read: September 22, 2026, re-checked October 20, 2026, with the same results—live credit quotes and the models' parameter contracts in LUVI.
• Results: 360 credits to 1080p and 480 to 2K for the ten-second clip. Wan 3.0 at five seconds: 400 credits at 480p, 1,600 at 1080p. PixVerse v6 at five seconds: 350 at 360p, 1,150 at 1080p.
• Drawbacks: we compared prices, not pictures. No outputs were generated for this post, so it makes no claim about how an upscaled draft compares with a native render at the same resolution.

In LUVI

Upload or pick the clip, open the Video Upscaler, choose 1080p or 2K, and read the estimate with the video already attached. If you are planning a batch of shots, the cheap-draft habit is the one that saves the most: render everything at the bottom tier, choose, then lift only what survives. The help center explains how resolution and duration change the price on the generating side of that decision.

New accounts start with 200 credits, and verifying a phone and a card adds up to 1,500 more. You can create a LUVI account and test the two routes on your own footage.