Case 02 / Automation
Exosoft video pipeline
AI video production in n8n, built with Brody Automates
- Role
- Automation developer, working with Brody Automates for Exosoft
- Year
- 2026
- Stack
- n8n, OpenAI, VEED
- Result
- 3h+ to 5 minproduction time per video
The workflow
- TriggerTopic in
- LLMWrite the script
- VoiceGenerate the audio
- VEEDProduce the video
- DeliverSend the file
- LogLog the run
An n8n workflow that turns a topic into a finished, voiced video, delivers it and logs the run. Production time went from more than three hours to under five minutes per video.
The problem
Every short video meant writing a script, recording or generating a voiceover, producing the video and sending it on. More than three hours of hands-on work per video.
What I built
- A trigger that takes a topic
- Script generation with an LLM
- Audio generation for the voiceover
- Video production through VEED
- Delivery of the finished file
- Retries on the steps that can fail, and an alert when a run still breaks
- A log entry for every run, so nothing happens invisibly
The hard part
VEED's n8n node had come out only days before, and Exosoft got early access to test it. There was no documentation yet, nothing useful on Google, and no AI model knew the node existed.
Every other step worked on the first pass. The VEED step did not. I spent two to three days testing it call by call, working out what it expected from what it returned, until it ran reliably. At one point the call was that it probably could not be made to work. It ships today.
The result
Production time went from 3+ hours to under 5 minutes per video.
