How we automated video editing and publishing at ZaTek

INTERNAL CASE · REEL STUDIO

We built Reel Studio to bring video preparation, editorial variants and social publishing into one workflow. This case covers our ZaTek and ZaTek AI installations.

The workflow starts with a video file. The system analyzes the material, prepares titles and captions, generates versions and retains cases that require review. Publishing has its own state and only counts as completed when a destination receipt exists.

Jobs created from September 28 through October 25, 2026, using Argentina time. States checked on October 26. Eight jobs marked as tests were excluded.

What the system recorded during that period

26 distinct source files. With at least one completed version. File fingerprints identify the originals without counting their variants as additional source files.

55 completed video jobs. From a cohort of 70 jobs. This includes originals, variants and reuse, rather than 55 distinct original videos. The remaining states were six awaiting review, five failed and four cancelled.

28 confirmed Instagram publications. Linked to those jobs and published within the same period. The ZaTek installation accounts for 12 and ZaTek AI for 16. Each publication is counted once per account, platform and identifier.

Publishing is measured separately from editing. There are 48 publishing records linked to the cohort: 28 completed and 20 in a failed state. A completed version does not mean it was published.

What this demonstrates and what we have not measured

The records support completed versions and publications with receipts. We do not have a prior measurement of manual working time, so this case makes no claim about hours saved, financial return, reach or sales. Receipts also do not establish that every publication remains available today.

The exclusion uses saved test flags and their source files. We cannot establish that every historical test was marked.

The workflow makes it possible to distinguish completed versions, published posts and cases that need intervention. Keeping those states separate helps identify an interruption without treating it as a completed delivery.

Method
Read only job and publication receipt records from the ZaTek and ZaTek AI installations. Cohort selected by job creation date, with states checked on October 26, 2026. Source files deduplicated by SHA-256. Publications linked by exact job identifier, completed state, media identifier, link and publication date within the period. Marked tests and identifiable derivatives excluded. No conversations or client data were consulted.

https://www.za-tek.com/en/case-studies#automatizacion-videos-zatek