How to budget your generation credits per scene before you generate a single shot
Most solo AI filmmakers spend their credits on the first scenes they feel excited about and ration the last ones. Here is a one-sheet budget that fixes the split before you start.
You start a film with a pile of credits and a feeling that it will be enough. The opening scene is the one you have pictured longest, so you generate it eleven times, chase a slightly better face, then a slightly better sky. By the time you reach scene four you are looking at a balance that covers one attempt per shot, and the back half of the film is made of whatever came out first.
The damage is not just visual. The scenes you rushed are usually the ones carrying your story turns, because the opening is easy to love and the middle is where plot lives. A film where the first ninety seconds are polished and the rest is first-take material reads as unfinished, however good the idea was. The fix is boring and takes twenty minutes: decide what each scene is allowed to cost before you generate anything.
Count shots, then set a take allowance
Open a spreadsheet or a plain text file. List every scene, then every shot inside it from your shot list. Next to each shot, write a take allowance. Use three tiers:
- Hero shots (the ones the scene cannot work without): 6 takes.
- Standard shots (coverage that needs to be clean but not special): 3 takes.
- Filler shots (inserts, cutaways, texture): 2 takes.
Multiply each tier by the credit cost of one generation on the model and settings you plan to use, then add it up. If you use different models or resolutions for different shots, price each shot at its own rate. Check your cost per generation in the tool itself, because it changes with duration and resolution and you should not guess it.
Hold back a reserve
Before you divide anything, take 20 per cent of your total balance and park it. Name it the reserve. It pays for reshoots after the rough cut, when you discover a missing reaction shot or a bridge between two scenes. Without it, every late discovery becomes a choice between a gap in the film and an overdraft on something you had already finished.
Your working sheet should look like this:
```
TOTAL CREDITS: 1000
RESERVE (20%): 200
AVAILABLE: 800
SCENE 1 | 6 shots | 1 hero, 3 standard, 2 filler | 6+9+4 = 19 takes
SCENE 2 | 8 shots | 2 hero, 4 standard, 2 filler | 12+12+4 = 28 takes
SCENE 3 | ...
TAKES PLANNED: 112 x cost per take = 784
SPENT SO FAR: 0
```
Add a spent column per scene and update it at the end of every session, not the end of the project.
Cheap tests first, expensive takes last
If your tool offers a shorter duration or lower resolution at a lower price, use it to test the prompt and the framing. Spend full-price takes only once the cheap version has the composition and motion you want. Log the test takes separately so they do not eat your hero allowance.
Stop rules
Write one rule at the top of the sheet: when a shot has used its allowance, you either pick the best take and cut around it, or you move on and return only if the reserve allows. This is the hard part, because the sixth take always feels like the good one. The rule exists so that feeling does not get to spend your scene five budget.
What you should see
Before: scene one has forty takes logged, the finale has two, and you discover the imbalance at the end. After: every scene has a spent figure within a few takes of its plan, the finale has hero-level attention, and you still hold a reserve for the two shots the edit asks for. The film reads as one consistent piece because every scene got the same discipline.
Try it on your current project tonight. List the scenes, assign the tiers, park the reserve, and see whether the numbers fit before you generate another frame.