Blog · October 6, 2026

Captions for video ads: style, timing and placement

Captions do more in a video ad than help people watching with the sound off. They hold attention, make the line easier to follow, and make the ad accessible to people who can't hear it. Done badly, they cover the product, lag behind the voice or misspell your brand name. This guide covers how to do them well: timing, style, placement and checking.

Burned in, or the platform's own?

TikTok, Instagram and YouTube can all generate captions automatically, but viewers can switch them off, the style differs from app to app and the quality varies. Captions burned into the video look the same everywhere and are part of the design. For ads, burn them in, and leave the platform's automatic captions off for that video so the words don't appear twice.

Timing: word by word or phrase by phrase

Captions should appear as the words are said, not a sentence at a time ahead of the voice. Two styles work well on short video:

• Phrase: each word pops in as it's spoken and stays, building the phrase up a few words at a time, then the next phrase replaces it. Easy to read, and it keeps the whole thought on screen.
• Word by word: only the word being said is shown, each one replacing the last. Punchy and attention-grabbing, best for short lines and fast delivery.

Either way, keep to about five words on screen at once. A full sentence in a block is too much to read on a phone while also watching the video.

Style

• A bold sans-serif font. Thin or decorative type disappears on a small screen. Condensed display fonts work for single words; rounder bold fonts for phrases.
• Big enough to read at arm's length, but not so big that a phrase takes over half the frame.
• High contrast. White text with a dark outline or shadow stays readable over any background.
• One or two lines at most.
• Consistent. Pick a style per brand or campaign and keep it, so the captions look designed rather than accidental.

Placement

Keep captions in the safe area, clear of the account name and caption text along the bottom of the app and the buttons down the right edge. Roughly a third of the way up from the bottom works on most videos. Then check two things on your actual video: the captions mustn't cover the product, and they mustn't cover the speaker's mouth. Move them up or down for the shot rather than forcing one position. How to make video ads for TikTok, Reels and Shorts covers the safe zone in more detail.

Don't ask the video model for text

Video models are poor at writing legible text; ask one for on-screen words and you'll usually get something that looks like letters but isn't. Generate the video with no text in it, list on-screen text and subtitles in the exclusions, and add captions afterwards from the real audio. More on keeping AI video clean in how to make an AI video that looks real.

Check every word

Automatic captions are made by transcribing the speech, and transcription can mishear brand names, product names and numbers. Before posting, watch the captioned version through once, checking:

• the brand and product names are spelled right,
• numbers, sizes and prices match your facts,
• each word appears with the voice, not before or after it,
• nothing important is hidden under the app's buttons.

Short lines make better captions

A line of 8 to 15 words makes two or three clean phrases on screen. A line of 30 words makes a wall of text that changes too fast to read. If the captions feel crowded, the script is too long; the UGC ad script template has a word budget for each length.

Captions in fairytale video

Adding captions to a finished video in fairytale video uses no extra credits. On a finished Seedance video, choose Add captions and the studio listens to the video and times each word as it's said. You choose one of nine fonts, from Montserrat and Poppins to Bebas Neue and Anton, and an animation (Phrase, which builds each phrase word by word, or Word by word). Then you place and size the text box on a live preview of the video, drawn with the same fonts as the final file. Render captioned video burns them into a new H.264 MP4, and the original video stays as it was. The captions show the words as the studio heard them, so watch the captioned video through before you post it.