How to make a YouTube Short with no source video — just a topic

Usually a Short is something you film first and cut later. But what if there's nothing to film — no footage, no source clip, just a topic? Vertix has a separate path for exactly that: topic mode. You send a topic as text, and the bot writes a script, voices it, and assembles a finished vertical video — without a single second of your own footage.

First, the bot figures out what kind of topic this actually is

Before writing a single word, the bot classifies the topic: is it a general explanation of something external, an inner feeling, a breakdown of a specific work, a story about a real person or event, pure fiction, or a how-to. The topic's type sets the rules the script has to follow — topics about real people and events, in particular, run under stricter honesty rules. That's its own story, covered in a companion post about why AI shorts generators sometimes lie about real people — and how that's fixed here.

You hear the script, not read it — and approve it by ear

The written script isn't shown to you as text for approval — the bot voices it immediately and sends it as a voice message. The logic: the finished video will be watched, not read, so it should be judged the same way, by ear. Under the voice message sit four buttons: "Looks good," "Rephrase it," "Edit the script," and "Cancel." You can ask for a rewrite up to three times if the delivery isn't landing, or, if you want to fix specific wording, send your own text and the bot reads exactly that, with no model call at all.

It's also cheap by design: the approved voiceover ships into the final render as-is, never re-synthesized. You pay for the voiceover once, and the finished video says exactly what you already approved.

The visuals are picked to match the text

Next comes the background. By default the bot pulls stock footage matching the meaning of the approved script — it sends four candidates with previews you can play before deciding. If none of them fit, you can send your own video and it becomes the background instead. If you build topic Shorts regularly, videos you've sent before are saved to a personal library and offered again later, so you don't have to hunt one down and resend it.

There's a third option too — a visualization of the voiceover's own soundwave instead of any video at all. It's genuinely, literally unique: two scripts that read differently physically cannot produce the same background, because the background is the shape of that specific voiceover.

Length is a target, not a promise

The bot doesn't promise an exact length — it's whatever the model actually ended up narrating, not something fixed in advance. The "Short length" setting sets a ceiling the bot aims for, but if the voiceover runs long, you'll see it before assembly even starts: the voice message gets an honest warning about how long it came out and what won't fit, with the same buttons to rephrase or fix the text yourself.

What comes out is already publish-ready

A finished topic Short always arrives with a YouTube title and description — that's not a toggle, it's baked into the build, because a video with no title is pointless on YouTube anyway. Words marked for emphasis in the script itself flash as large on-screen text in time with the narration — a separate layer from the subtitles, always on, not an option you have to remember to enable.

If this isn't your first topic Short

The bot remembers your past topics — once you've taken at least one topic Short all the way to the render queue, a "Suggest a topic" button appears next to the input field: the bot looks at your recent topics and proposes a few new ones in the same area, without repeating itself.

All of it, with zero seconds of source footage. Open the bot, send a topic as text — and listen to what comes out before you build anything.