Making Vertical AI Clips for Reels and TikTok (And What You Cannot Post)
By The Fellowi Team · · 5 min read
Vertical is a different craft from landscape, and most of what makes an AI clip look good in a 16:9 frame quietly stops working when you turn it on its side. A wide establishing shot becomes a strip of nothing. A slow push that reads as cinematic at full width reads as a frozen frame on a phone. The format punishes patience.
Fellowi Video does 9:16 natively and up to 30 seconds, so the tooling is not the constraint. What follows is the practical part, plus one thing about publishing that people find out the hard way.
Set the ratio before you generate, not after
Cropping a 16:9 clip down to 9:16 afterwards throws away most of the frame and usually cuts the subject in half. Choose 9:16 in the generator instead, and the model composes for that shape from the start.
If you already have a vertical start image, the better option is often auto, which keeps your image's own framing untouched. Forcing an explicit ratio onto a still that already has one is how you get unexpected padding. The full set of options and what each costs is in what Fellowi Video can actually do.
The first second is the whole decision
On a vertical feed, viewers decide in well under a second. That has a direct consequence for the prompt: put the movement at the start, not at the end. “She turns toward the camera and smiles” is a first frame worth stopping on. “She stands still, then after a moment turns” spends your only chance on stillness.
This is also why four to eight seconds tends to outperform thirty. A short clip that loops cleanly gets watched three times; a long one gets abandoned at second four. Pay for length only when the shot genuinely needs it, and remember that you are charged per second. The general motion vocabulary is in how to write AI video prompts.
Sound: usually off
Clips can carry a generated soundtrack matched to the scene, free of charge. For social, turn it off more often than not. Instagram and TikTok both reward using a track from their own audio library, and a generated soundtrack underneath it is just something you will be muting in the editor anyway.
Keep it on when the clip is the finished thing: a standalone post, a message, something you are sending rather than publishing.
Two stills beat one prompt
The feature that changes vertical work most is the end frame. Give the clip a starting image and a finishing image, and the model builds the motion between them instead of improvising where the movement lands. For a short vertical shot, where you have almost no room for drift, that control is worth more than any extra adjective. Combine it with keeping a character consistent and you can carry the same person across a series of clips.
The part nobody tells you
Fellowi allows adult generation. Instagram and TikTok do not allow adult posting, no matter how the content was made. Those two facts are not in conflict, but people conflate them and then lose an account over it.
So be clear with yourself about which job a clip is for. The generator is the same tool in both cases; the difference is where the file ends up. If you are making content for a public feed, make it publishable. If you are making something adult, it is yours, it is not watermarked, nobody reviews it, and it is not for your grid. Our own limits are in the uncensored video policy, and they are about what we will render, not about what a social platform will host.
One more practical note: label AI content when the platform asks for it. Several now detect it from file metadata anyway, and an unlabelled clip can quietly get demoted.
Ready? Open the Video tab, set 9:16, and put the movement in the first second. If you want the still first, generate it vertical so auto framing does the rest.