How many AI video generations do you need to budget for a short film?
Last updated June 26, 2026
Budget roughly 3 video generations per usable shot and expect about 25% of your clips to reach the final cut. Documented productions ran 164 generations for a 3-minute animated film and ~400 for a 90-second short — roughly 55–270 generations per finished minute, depending on style and selection discipline.
Start your budget from two documented ratios: an average of 3 generations per usable shot, and a ~25% clip selection rate from raw generations to final timeline. In one 3-minute animated production, 164 clips were generated and 41 made the cut, with an average of only 5 seconds used from each 15-second clip. Treat that overgeneration as a deliberate budget line, not waste — editorial selection is where the quality comes from.
Documented generation counts and costs. Two productions logged exact generation totals, and they show how wide the legitimate range is:
| Production | Finished length | Video generations | Made final cut | Total cost |
|---|---|---|---|---|
| Animated episode (2-person team, 2 days) | 3 min | 164 | 41 (~25%) | ~$950 |
| Horror short (2 days) | ~90 sec | ~400 (+30 image gens) | — | $870 (4,100 credits) |
That works out to roughly 55–270 video generations per finished minute. Across five documented productions, total spend ran $750–$5,000, or $315–$750 per finished minute — different teams and styles legitimately land at different points in that range.
How to estimate your own number. Build a shot list first, then multiply shot count by 3 generations. Add character lock-in on top: one production needed about 5 generations to lock each character's identity, at roughly $9.78 per character, plus 11–30 image generations for reference sheets and props across documented projects. Then discount for harvesting — each 15-second clip typically contained 4–7 usable shot candidates, so one generation can feed several timeline shots if you cut selectively in your film's aspect ratio and delivery format.
How to keep the count down. invideo is an agentic video creation tool with all the current video models — Veo, Kling, Seedance 2.0 — available in one place, and three of its controls directly cut wasted generations. First, lock character sheets and environment references before any video generation: one production generated 4 options per asset, picked one, and locked it, which prevents consistency re-rolls later. Second, run the invideo agent in Always Ask mode so you approve every prompt and attached reference before credits are spent. Third, let the invideo agent flag model limitations up front — in one production it caught that an 18-cuts-in-15-seconds scene exceeded what the model could deliver and recommended splitting it before any credits were burned. During the edit, stitch keepers instead of chasing a perfect single take: 17 of the 41 final shots in the animated episode — over 40% — were Frankenstein shots assembled from 2 or more generations of the same prompt. And when a continuity error appears, fix it at the character-sheet source through the invideo agent rather than re-generating the shot, which keeps the rest of your generation budget intact.
These ratios shift with your style and shot complexity, so treat 3 generations per shot and a 25% keep rate as the planning floor, then adjust after your first scene.
Watch some of these to see what works for you:
Out of 164, 41 videos made the cut, and on average only 5 seconds of each 15-second clip was used. That's how 41 clips became a 3-minute episode.
— invideo's creative team