Reuse Is A Search, Not A Memory

On 11 September 2026, at 16:25, the scan ran and the stamp was written to the minute. 303 assets across 12 projects went in. 226 came back reusable. 77 were excluded as archive, frames or rejects. The number that runs the studio is not 303. A library of 226 usable clips is a supply. A folder of 303 is a hope. The engine reports what it can cut from, not what it holds, and that single distinction is why the index exists at all. 1 pass, 1 stamp, 2 counts. The probability that a team re-renders a clip it already owns climbs every month the index goes unsorted, because nobody can search a memory. Sorting is not the job. Searching is the job. The count is only the proof that a search will return something.

Reusable assets per project tell you nothing without footage. Sal_y_Miel holds 147 reusable assets and 24.5 minutes. TikTok Short-Film holds 28 and 3.4 minutes. H3_Dialogue_Test holds 0 reusable assets against 2.9 minutes, which is information rather than failure: those 12 clips are tests, and a test does not go in a cut. Salvador holds 10 assets and 6.0 minutes. Apocalypse holds 15 and 1.4 minutes. Songs holds 8 and 3.4 minutes. Lola holds 9 and 0.0 minutes. Anna Meu Amor holds 3 and 0.0 minutes. 2 Nolan folders hold 2 assets each and 0.3 minutes each. Counted by project and by minutes, the library stops being 1 number and becomes a map of what each project can already pay for.

The same 226 assets answer a second question before they answer anything else. 86 are landscape clips at 16 by 9. 18 are portrait clips at 9 by 16. A vertical deliverable cannot be served by a horizontal clip, so the portrait count is its own budget: 18 shots available to every vertical film the studio takes on, against 86 for every horizontal one. Coverage is only honest when it is counted per shape, per project and per minute. 3 cuts of the same total give 3 different answers, and the 1 that matters is the answer to the question the next job actually asks. A total is a headline. A shape count is a schedule.

2 near-duplicate clusters were found and excluded: 1 at 10.12 seconds and 1366x768, 1 at 5.88 seconds and 1366x768, each present twice under 2 folder spellings. The same shot was indexed twice because the project directory was written 2 ways, with the hyphen and without it. Counting 2 copies of 1 clip as 2 clips is how a library starts lying about itself, and the lie is always in the same direction: it overstates what the studio owns. 2 rows removed, 1 honest number left. Deduplication is the quiet part of the job and the part that decides whether the count can be trusted for a delivery date.

303 assets in, 226 usable, 77 excluded. 1 of those 3 numbers runs the studio.

The reuse-first rule sits at the top of the file, above every row of the catalog: before rendering anything new, run 1 search for the subject, and if a reusable clip matches the beat it goes in the cut. 1 command, written first, before the coverage tables and before the near-duplicate log. It runs before the render, not after it. Put the search after the render and the folder fills with clips nobody can tell apart 3 months later. Put it before the render and the same folder behaves like a stock library with 1 owner. 1 search, 1 scan, 2 orders of work. The difference is only which 1 runs first, and the bill for getting that wrong is never the render fee. It is the version confusion that follows the render.

The pipeline that carries EX Venture's studio output runs on the same test: 35 staff and 200+ interns a year, every render order passing the same 1 question first, and every render order answering to the same index first. The index is not a gallery. It is the reason the next render can be refused, and refusing 1 render is worth more than 10 cleanups after it. A stranger cannot check a folder by opening it. A stranger can check an index in 1 command, which is the difference between a studio with a supply and a studio with a story about a supply. Search, then order. 226 usable assets, 86 landscape, 18 portrait, 1 rule, 0 arguments.