Skip to main content

​
The Reframing Problem

When you crop a 16:9 master to 9:16 for Stories, you lose 75% of the frame. A subject centered in widescreen might be cut off entirely in portrait. Manually repositioning the crop for every shot across every format is tedious and expensive.

​
How Smart Reframe Works

Versionary uses Google Video Intelligence API with person detection to analyze every frame of your source video. The system:
  1. Detects shot boundaries — identifies every cut in your footage
  2. Tracks subject position — locates the primary subject’s face and body position in each shot using nose landmark tracking
  3. Calculates per-shot crops — positions the crop window to keep the subject centered for each target aspect ratio
  4. Builds dynamic expressions — creates frame-accurate crop positioning that changes at every shot boundary
The result: each aspect ratio gets its own intelligent crop that follows your subject through the entire spot, changing position at every cut.

​
Supported Aspect Ratios

RatioCommon UseCrop from 16:9 Source
16:9YouTube, Broadcast, OLVFull frame (no crop)
9:16Instagram Stories, TikTok, Reels56% of width preserved
4:5Meta Feed, Instagram Feed75% of width preserved
1:1LinkedIn, Square placements56% of width preserved

​
Accuracy

Smart Reframe achieves approximately 75% accuracy on initial auto-crop positioning. For shots where the AI doesn’t nail the framing — action sequences, wide shots with multiple subjects, or artistic compositions — the conversational editing system lets you make instant corrections.

​
Under the Hood

The system builds nested FFmpeg conditional expressions that evaluate at each shot boundary: 
crop=608:1080:if(lt(t,1.5),420,if(lt(t,3.2),538,if(lt(t,5.0),612,420)))
This single expression tells FFmpeg to position the crop at x=420 for the first shot, x=538 for the second, x=612 for the third, and so on — all in one render pass with zero quality loss.