How to make vertical videos for TikTok, Reels and Shorts

Choose the frame before anything is generated, and you get a real vertical video — not a landscape one letterboxed into a strip. Here is how the shape, the captions and the cost actually work.

Made with the app, in the format this guide is about: 9:16, captions lifted clear of the platform’s own buttons.

Pick the frame before you generate, not after

The single thing that separates a vertical video that works from one that does not is choosing the shape up front. A 16:9 video posted as a Short or a Reel gets letterboxed — black bars top and bottom, your subject shrunk into a strip in the middle. Cropping it afterwards cuts the sides off every shot.

So the frame is a choice you make before anything is generated:

Every scene is composed at that shape from the start. Stock footage that does not match is filled rather than letterboxed — the clip sits over a blurred, cover-cropped copy of itself, the convention every vertical-video app uses, instead of stranding a thin strip in black.

Captions sit higher in vertical, on purpose

TikTok, Reels and Shorts all put their own interface along the bottom of the screen — the caption, the handle, the buttons. Subtitles positioned for landscape end up underneath all of it. In 9:16 the burned-in captions are raised so the platform's own furniture does not cover them.

This matters more than it sounds: most short-form video is watched muted. If the captions are hidden, the video says nothing.

Or say it to Claude, with the shape in the sentence

The app connects to Claude as an MCP connector — one URL, added once under Settings → Connectors → Add custom connector:

Because the frame is fixed at creation and cannot be changed afterwards, put it in the request rather than hoping for the default:

Claude asks what it needs, sets 9:16 before anything is generated, and renders it. If a scene misses, ask it to drop that scene — the re-cut reuses the clips already made and costs nothing. How the Claude connector works →

Making one

1. Type a topic, or paste a script if you already have the words.
2. Set Frame to Shorts, TikTok, Reels · 9:16.
3. Pick where the pictures come from — free stock footage, free stock images, or AI stills generated on your own connected Pollen.
4. Choose a length. Around 40 words of narration is roughly 30 seconds.
5. Generate, then review it scene by scene before anything is published.

Cut it down without regenerating anything

Short-form punishes slow openings. Every scene is stored separately, so dropping the weak ones is a re-stitch of clips that already exist: it takes seconds, costs nothing, regenerates nothing, and you can put a scene back afterwards if you change your mind.

That means the honest workflow for vertical is: generate a little long, watch it once, drop the two scenes that drag, re-cut. No timeline, no re-render, no extra spend.

What a vertical video costs

The frame itself is free. What decides cost is where the pictures come from:

• Free stock footage and free stock images — real media, unlimited, no Pollen spent, no subscription.
• AI stills — one generation per scene, on your own connected Pollinations account, so you control the spend.

You can also set a daily limit on AI generations that falls back to free stock once it is reached, so a runaway series cannot empty your balance.

Next: turn a script you have already written into a video, or ask Claude to make it for you.