clipsdk turns a wide 16:9 recording into vertical 9:16 clips by tracking faces and keeping the person who is talking in frame. Pick auto, center or none, and an output of 9:16, 1:1, 4:5 or 16:9, per request.
Pick an output size and a mode. The red window on the 16:9 recording is the crop, and the clip on the right is what you get back.
aspect
reframe
Who is talking
9:16
Source: 16:9. The red window is the part that becomes the clip. Click a face to change who is talking.
Onecoldemail.
9:161080x1920
Auto: Tracks the faces in each shot and keeps whoever is talking in frame. It cuts when the speaker changes, holds through short interruptions and frames the center when nobody is on screen.
{"aspect": "9:16","reframe": "auto"}
Auto mode
It follows the conversation
Here is what auto does over 40 seconds of a two-person podcast.
Talking
MayaDanielMayaDanielMaya: "Right."
Crop
0:000:100:200:300:40
Cut when the speaker changesNo cut for a one-word reply
1Finds the facesEach shot is checked for faces, so the crop always knows where the people are.
2Follows the voiceThe crop stays on whoever is talking, so a guest's answer is framed on the guest.
3Cuts, then holdsWhen the speaker changes, the crop cuts instead of panning. A one-word reply doesn't trigger a cut.
Modes
Auto, center or none
One field on the request. Here is the same moment made three ways.
reframe: "auto"
Tracks the faces in each shot and keeps whoever is talking in frame. It cuts when the speaker changes, holds through short interruptions and frames the center when nobody is on screen.
reframe: "center"
A fixed crop from the middle of the frame. Good for one speaker who stays put.
reframe: "none"
No crop. The full 16:9 frame sits inside the output size on a blurred fill. Good for slides, screen shares and wide shots.
Which mode for which recording
Recording
reframe
aspect
Two-person podcast or interviewCuts to whoever answers
"auto"
9:16
Panel or roundtableStays on the voice, wherever they sit
"auto"
9:16
Keynote or talk on a stageKeeps the speaker in frame
"auto"
9:16
Solo talking head or webcamOne face that stays put needs no tracking
"center"
9:16 or 4:5
Slides, demos or screen sharesNothing gets cut off at the edges
"none"
16:9 or 1:1
Wide shot of a groupEveryone stays in the picture
"none"
1:1 or 4:5
Output sizes
Four shapes, all at 1080p
Set aspect once per request. Every clip in the job comes back in that shape.
9:16
1080x1920
Shorts, Reels and TikTok
1:1
1080x1080
Square feed posts and ads
4:5
1080x1350
Instagram and LinkedIn feeds
16:9
1920x1080
YouTube and the web
Recording tips
Get a better crop from the same camera
Film everyone in one shot
For a podcast or interview, send the episode with everyone in frame. The crop does the switching, so a two-shot becomes a close-up of whoever talks.
Follow the speaker on stage
For talks, a camera that follows the speaker works better than a wide shot of the whole hall, where the speaker is small.
Big game, small facecam
The 9:16 frame centers on your face. If most of the picture is gameplay, ask for 16:9 to keep all of it, or 1:1 for a middle ground.
Slides that must stay readable
When the slide matters more than the face, use none. The whole frame stays in the clip on a blurred fill.
FAQ
Questions, answered
How does auto reframe decide who to follow?
Tracks the faces in each shot and keeps whoever is talking in frame. It cuts when the speaker changes, holds through short interruptions and frames the center when nobody is on screen. There is nothing to tune: set reframe to auto.
What output sizes can I get?
9:16 at 1080x1920, 1:1 at 1080x1080, 4:5 at 1080x1350 and 16:9 at 1920x1080. Set aspect on the request and every clip in that job uses it.
What does reframe none do?
No crop. The full 16:9 frame sits inside the output size on a blurred fill. Good for slides, screen shares and wide shots.
When should I use center instead of auto?
When one person sits in the middle of the frame the whole time, like a talking-head video or a webcam recording. Center keeps the crop still, so the clip never cuts.
Does it work with slides and screen shares?
Use none to keep the whole frame readable on a blurred fill. Auto frames the center when nobody is on screen, which can cut off the sides of a slide.
Does reframing cost extra?
No. You pay $0.03 per minute of source video, and every clip comes back reframed and captioned at that price.
