Lesson 3 — The credit math: what actually costs you
Module 1 · ~12 min · platform data read on 2026-10-21
Credits are the only genuinely limited thing on this platform, and almost everyone learns to save them after having spent them. This lesson reverses the order.
The three things that move the price
Resolution. Cinema Studio 3.0's own documentation says it, right in the resolution parameter: "higher = more credits". The options are 480p, 720p, 1080p and 4K, and the default is 720p. Which means the default is already not the floor — if you've never touched it, you're generating above the cheapest setting.
Duration. Cinema Studio 3.0 runs 4 to 15 seconds; v2 runs 3 to 12; Seedance 2.5 reaches 30. A second more is a bigger generation.
The model. They don't cost the same, and the most explicit case is FLUX 3 Video Edit, whose description reads: "costs 1 credit per second of the processed clip" — and it uses at most the first 15 seconds. So sending a ten-minute video to be edited doesn't cost ten minutes, but it also doesn't use the whole video.
The testing rule
The saving isn't in generating less. It's in generating cheap while you're still searching, and expensive only once you've found it.
In practice:
- Test at 480p and the minimum duration. You're checking framing, motion, and whether the model understood — none of that needs resolution.
- Once the take is right, generate it again at 1080p or 4K, with the same prompt and the same start image.
- The expensive version is one, at the end. Not ten, in the middle.
People who do the opposite — testing in 4K because "might as well make it good" — spend the whole month on five attempts.
The switch nobody checks
Audio. On Cinema Studio 3.0 it comes off by default. On v2 and on Seedance, it comes on. If you're going to add music and voiceover in the edit, you're generating audio you'll throw away — on every take.
Turning sound off while testing is the easiest saving in this lesson, and you can
do it right now: on the models that expose it, that's sound: off or
generate_audio: false.
The math that closes the module
Before starting a project, run this one-line estimate:
test takes × cheap cost + final takes × expensive cost
If you're planning six cuts in a video and you usually need three attempts per cut, that's 18 tests + 6 finals — not 6 generations. That's the number that decides whether the project fits your balance, and it's the one nobody runs before starting.