What it does
The pipeline fetches current news, has Gemini write image prompts and copy, generates images, music and a voice-over, and assembles a vertical video with ffmpeg. At the end an MP4, subtitles and accompanying texts are ready.
Highlights
- Gemini creates four image prompts plus a hook and caption
- Images from IONOS FLUX.1-schnell via cloud API (no local GPU needed)
- Music with ACE-Step, local on Apple Silicon; voice-over with Piper
- Voice and music mixed with ducking, video in 9:16 (1080×1920) with Ken Burns motion and captions
- One run took about 10 minutes in testing on an older MacBook Pro
Tech
Python Gemini FLUX ACE-Step ffmpeg
Status
Working prototype, public on GitHub. Last updated in July 2026.
Source code
The source code is public on GitHub: github.com/Mikweb2025-design/zimage-turbo-pipeline