Short-form AI video in 2026 is being shaped by five typography moves that have moved from experimental to default: word-by-word kinetic captions with keyword colour pops, oversized centered hooks in heavy sans-serifs, hand-written or marker-style annotation layers, mixed-case bold for emphasis instead of all-caps, and animated text that scales or rotates in time with voiceover beats. These styles started in creator content, got absorbed into performance marketing, and are now baked into the default templates of nearly every AI video tool on the market. The trick is knowing which ones still convert and which ones are starting to feel like 2024.<\/p>\n\n\n\n
Typography in this format is doing real performance work, not decoration. Sound-off viewing rates on TikTok and Reels sit somewhere between 70 and 85 percent, so the text on screen is carrying most of the message. When the typography is right, it lifts hook rate and watch time meaningfully. When it’s wrong or dated, the viewer reads “this is an ad” in the first half-second and scrolls before the script even starts.<\/p>\n\n\n\n
Word-by-word caption animation, where each word or short phrase appears synced to the spoken audio, is the format that defined the last two years of short-form video and isn’t going anywhere. The original push came from CapCut’s auto-captions and the rise of creators like Alex Hormozi, who burned a specific look (chunky Montserrat, yellow keyword highlights, slight bounce on entry) into the cultural memory of the format. By 2025, almost every AI video tool shipped this style as a one-click preset.<\/p>\n\n\n\n
What’s shifting in 2026 is the pacing. Early kinetic captions cut between words too fast, often at 200 to 240 words per minute equivalent, which created a sense of urgency but also a sense of cheapness. The current move is slower and more deliberate, with two to four-word chunks held for 600 to 900 milliseconds each. That gives the eye time to actually read the word before it disappears, which research on caption comprehension has consistently linked to higher information retention. The aesthetic still feels modern, but it stops triggering the scroll reflex.<\/p>\n\n\n\n
The first three seconds of a short-form video used to feature a small caption at the bottom of the screen. That convention is gone. The current format puts the hook in the dead centre of the frame, in a heavy weight, often at 80 to 120 pixels equivalent on a 1080×1920 canvas, with the actor or product framed behind or around it.<\/p>\n\n\n\n
The reasoning is simple. Users scroll on autopilot, and the centre of the screen is the only place that reliably interrupts the scroll. A hook reading “I tried this for 30 days” or “Don’t buy this until you watch” lands harder when it occupies the visual centre than when it’s tucked into a lower third. The fonts doing most of the heavy lifting here are TheBoldFont, Montserrat Black, Anton, Bebas Neue, and Druk Wide for more premium-feeling content. Mixed case (capitalising significant words rather than all-caps) has overtaken full caps for most categories because all-caps now reads as either shouty direct-response or visually dated.<\/p>\n\n\n\n
The opposite move, and an interesting counter-trend, is the rise of hand-drawn annotation layers over otherwise clean video. Yellow highlighter swipes under key phrases, red circles around products, scribbled arrows pointing at on-screen details. These are often added on top of standard captions, not in place of them.<\/p>\n\n\n\n
The reason they work is that they read as human attention. A perfectly rendered text overlay signals production. A slightly wobbly yellow highlight signals that a real person decided this part mattered. Even when the entire video is AI-generated, that one annotation layer pulls the perceived authenticity up several notches. The style works particularly well for tutorial content, product comparisons, and reaction-format ads, and it works less well for luxury or premium brands where the deliberate imperfection feels off-key.<\/p>\n\n\n\n
Most production teams aren’t choosing fonts from scratch any more. They’re choosing presets inside AI generation tools and tweaking a few variables. That changes how trends propagate, because whatever ships as the default in the leading platforms becomes the default look for thousands of brands within weeks.<\/p>\n\n\n\n