Automated content: Rooted_Wisdom
From topic to a scheduled long-form video: script, voice, images, edit and upload.
Running on a scheduleRuns on a schedule, producing finished videos with no manual editing.
What it does
Produces a complete long-form YouTube video from one topic and uploads it privately for review.
What I built
Claude writes the script, AI makes voice and images, FFmpeg assembles with captions, the YouTube API schedules it.
Result
Runs on a schedule on my server. The same system was reused for a second channel with source-grounded scripts.
How a video gets made
- System Topic from analytics
- AI Claude writes the script
- AI Voice and ~20 images
- System FFmpeg edit and captions
- Person Private upload for review
- System Scheduled publish, stats feed back
Technical details
- Every stage writes its state, so a failed run resumes where it stopped.
- For the second channel, a retrieval step pulls excerpts from official sources before the script is written.
- Image-generation approaches compared against real viewer retention.
- Uploads, scheduling and analytics through the official YouTube APIs with OAuth2.
Stack
- Python
- Claude Code CLI
- TTS
- AI image generation
- FFmpeg
- YouTube Data & Analytics API
- cron
Channel analytics stay private. The thumbnails are from videos published on the channel.