Head of Production for 5,000-Client Edtech — 4.5× Faster Pipeline
Built a 4-person content team from zero in a 20-person edtech company. Then automated the pipeline with AI to cut editing time from 3 hours to 40 minutes per lecture.
4.5×
Editing speedup600+
Lectures produced20
Courses delivered3-4/day
Lecture throughputThe Challenge
KCE needed to scale from a handful of courses to 20+ without growing the team proportionally. Each lecture took about 3 hours to edit — multi-cam sync, graphics, b-roll, QC. At 3-4 lectures per day, the math didn't work.
I was brought in to build the content production department from scratch. No team, no processes, no templates.
The Approach
Built a 4-person team and a multi-layer automation pipeline:
Layer 1 — Shoot optimization: Midjourney-generated thematic green screen backgrounds per course. No physical set changes. Multi-cam setup with 3 audio sources and prompter.
Layer 2 — B-roll pipeline: n8n workflow analyzes lecture transcripts, extracts key concepts, generates visuals via Midjourney, animates through Runway/Kling 3.0, outputs ready-to-edit b-roll.
Layer 3 — AI talking heads: HeyGen + ElevenLabs for supplementary presenter segments. NotebookLM for automated presentation generation.
Layer 4 — Post-production: ExtendScript batch processing in Premiere Pro across hundreds of lectures. ISO 9001:2015 certified QC process.
5 stages · fully automated
- 1Source VideoRaw lecture footage
- 2n8n AnalysisTranscript extraction & concept analysis
- 3AI Image GenMidjourney / DALL-E visual generation
- 4Hailuo AnimationRunway / Kling motion generation
- 5Premiere AssemblyExtendScript batch processing
Animated diagram showing data flowing through five pipeline stages: Source Video, n8n Analysis, AI Image Gen, Hailuo Animation, Premiere Assembly.
The Result
Editing time dropped from 3 hours to 40 minutes per lecture — a 4.5× speedup. 600+ lectures across 20 courses delivered at 3-4 per day. I also shot courses myself: multi-cam, green screen, full-cycle from recording to export. The team I built continues to operate on these processes.
