Daily Video Pipeline — One Vertical, Three Platforms, Five PM

The brief

Short-form video is the loudest organic surface and the slowest to make by hand. The wrong shape is a thirty-minute editing session per clip. The right shape is a pipeline that runs while the operator sleeps and ships a clip every day at the same time across three platforms — and a self-critique loop that catches the bad days before they post.

The clip has to be ours, not generic. A gorilla character, one stylistic mode per day, a biohazard logo top-right, a follow-to-smile outro CTA, ground-truth-aligned captions. The pipeline owns all of that.

What we built

A nine-stage pipeline. Concept generation pulls a daily spec — four still prompts, narration, mode, gorilla style, HyperFrames disruptions, outro CTA. Voice synthesis hits a timestamped endpoint and clamps speed to land between seven and a half and eight and a half seconds, then builds caption phrases from the alignment so the captions match the audio frame for frame. Stills generate sequentially with the first still passed as a reference to two through four, locking character continuity across the clip.

HyperFrames composes the four stills with Ken Burns motion, crossfades, and disruption effects — RGB-split, paper-flash, zoom-punch, red-flash. Three post-processors run after the render: a kinetic intro pass, audio stings on beat hits, word-pop captions synced to the timestamps. A self-critique pass scores the clip against a rubric with hook and smile double-weighted. Below six it regenerates from concept. Above seven and a half it ships.

Delivery hits X, LinkedIn, and Instagram through Buffer at five PM Eastern through an n8n schedule, with the operator scoring the clip back through a Telegram reply that feeds the next day’s concept.

The receipt

The clip ships every day. The verdict gate kills the bad ones before they reach an audience. The character is the same character — one gorilla, one daily style — across every clip. The audio cadence aligns with the captions because the captions come from the audio.

What this signals for a buyer: video as infrastructure rather than video as project.