This n8n workflow automates a YouTube video pipeline. It takes video ideas from a Google Sheets backlog and uses GPT-4o to write a 14-scene script. It then renders the video with JSON2Video and publishes it to YouTube only after a manual approval.
On a schedule, it fetches the first Google Sheets row whose PRODUCTION STATUS is "to create." The idea title, captions and style fields go to GPT-4o, which returns a JSON script. Each scene has two-sentence voiceover text and a cinematic image prompt. Music and intro assets are read from a second Sheets tab.
The script, voice settings (via ElevenLabs), image model, and audio and intro URLs are submitted to the JSON2Video API. The workflow waits, polls the render status, and repeats until the render completes or fails. On completion, it saves the movie URL to Sheets and marks the row as created. On failure, it logs an error status.
- Platform
- n8n workflow
- Script model
- OpenAI GPT-4o, 14-scene JSON script
- Voiceover
- Two sentences per scene, ElevenLabs voice settings
- Video rendering
- JSON2Video API via HTTP Request, API key credential
- Idea source
- Google Sheets row with PRODUCTION STATUS "to create"
- Music and intro
- Selected "final" assets from a second Google Sheets tab
- Status tracking
- Render URL and created, published or error status written to Google Sheets
- Publishing
- YouTube upload after PUBLISHING STATUS is set to "approved"
- Credentials
- Google Sheets OAuth2, YouTube OAuth2, OpenAI, JSON2Video API key
- Core nodes
- if, wait, switch, HTTP Request, YouTube; includes Sticky Notes