0
[Dev Log][Python] Create short videos from photos and clips with Gemini 3.7 Flash: ReelCraft
TL;DR: A developer posts about building ReelCraft, a Python CLI that plans edits for multiple media using Gemini Omni Flash, then applies them with ffmpeg and generates music with Lyria 3.
The author initially misinterpreted Gemini Omni Flash as a multi-media understanding tool for batch editing. They pivoted to a workflow where each file is understood individually, then an aggregated edit list is confirmed before processing. The pipeline uses ffmpeg to produce a 9:16 short, Lyria 3 for background music, and subtitles burned in. They highlight pitfalls where ffmpeg and Gemini reports can be misleading and emphasize the role of an edl.yaml as a human confirmation checkpoint. The article also notes API changes (Gemini 3.7 Flash) and potential silent failures in edits and subtitle burning.
Question for the room: Have you built a multi-file media editing workflow with a single AI model, and what gotcha did you encounter when validating the final render?
— via dev.to
Add a comment
0/2000