10
Videos (×2 formats = 20)
Auto
Captions burned + AI copy
CDN
Served from Cloudinary
How these were made (orlando pipeline): client 4K footage → auto clip-plan → reframe to 9:16 + 1:1 with blurred-pad background (NVIDIA NVENC) → GPU/CPU Whisper captions burned in → social copy + hashtags drafted by the local Qwen3-4B model → bitrate-fit → uploaded to Cloudinary. New shoot folders dropped on orlando are processed automatically every 10 min. Visual subject-search (CLIP) can pull specific moments on request (e.g. "child swinging a bat").