Research / Field Results
Testing Short-Form Video Automation at Scale
I built a local system to test how much of an Instagram and TikTok content workflow could be automated when the source library already contained premade short-form clips.
Published on
Instagram Peak
14 million views and more than 1,000 followers gained with this method.
TikTok Peak
More than 10 million views and more than 4,000 followers gained with this method.
The Experiment
Start with a library of premade clips, automate the repetitive production work, and measure how far a consistent publishing method can reach.
Why I Built It
I created Short-Form-Video-Automation to answer a practical question. How much of running TikTok and Instagram accounts could be automated when the raw material was already available as premade short-form clips?
The goal was not to generate every video from scratch. I wanted to build a repeatable local workflow that could organize clips, create and manage overlay text, match the right text to the right clip, render a finished vertical video, and prepare it for publishing.
How the System Works
The project uses a local Flask dashboard and SQLite database to manage clips, overlays, matches, renders, and schedules. It can import clip metadata from CSV files, generate overlay ideas with a local Ollama model, and rank clip and text combinations with deterministic matching.
FFmpeg turns the selected combinations into 1080 by 1920 videos with text overlays. The current source-available version can publish finished reels to Instagram through Cloudflare R2 and the Instagram Graph API. TikTok publishing remains manual, but the clip selection, text workflow, matching, and rendering can still be handled locally before upload.
Peak Results
At the peak of this method, the Instagram account reached 14 million views and gained more than 1,000 followers. The TikTok account reached more than 10 million views and gained more than 4,000 followers.
Those results showed me that a large part of short-form production can be systemized when the source clips are already prepared. The strongest advantage was consistency. Instead of rebuilding each post by hand, the workflow made it easier to reuse the media library, track what had been used, and keep finished videos moving.
What I Learned
Automation works best on the repetitive parts of the process. Organizing files, generating options, matching assets, rendering videos, and tracking schedules are all good candidates. Human judgment still matters for choosing the source material, setting the direction, reviewing the output, and understanding what an audience actually responds to.
This project became a practical test of media operations, not just a rendering tool. It helped me see where automation creates real leverage and where attention still needs to stay human.