A practical guide to caption font, size, line length, safe-zone placement and keyword emphasis for TikTok, Reels and Feed video watched with the sound off.
Most social video is watched with the sound off — on a commute, in a waiting room, scrolling next to someone sleeping — which means your captions aren't a nice-to-have accessibility feature, they're the primary way most viewers experience your ad's hook and message. Badly styled captions, ones that are too small, too long-lined, or sitting under a platform's UI, silently kill performance for videos with a perfectly good script and a perfectly good hook. This tutorial covers font weight and size that actually reads on a phone, line length, positioning against each platform's safe zones, emphasizing the words that matter, contrast, and the final check you should always run on an actual phone screen before publishing.
Takes 3 minutes
Use a bold, sans-serif font — thin or script fonts disappear against busy backgrounds and cost you readability exactly when you need it most, in the first two seconds of the hook. Size captions so the text takes up roughly 8 to 10% of the frame's height; anything smaller forces viewers to lean in, which almost nobody does for a scrolling feed video.
Test your caption size the way your audience will actually experience it: hold the phone at a normal arm's length, glance at the frame for one second, and check whether you can read the full line without focusing hard. If you have to concentrate to read it, it's too small — remember that most viewers are giving the video far less attention than you are while reviewing it.
Avoid mixing more than two font weights in a single video. One bold weight for standard captions and one heavier or larger weight reserved specifically for emphasized keywords is enough contrast to guide the eye; adding a third style, like an italic or a script accent font, usually just adds visual noise without adding clarity.
Takes 5 minutes
Break captions into short bursts of four to six words rather than displaying a full sentence at once. Each caption block should express one complete idea, timed to appear and disappear roughly in sync with when those words are spoken, so the text never lags behind or races ahead of the audio.
Short caption bursts do two things longer captions can't: they keep pace with natural speech rhythm, and they create a visual pulse that helps hold attention frame to frame, since new text appearing is itself a small motion cue that competes for attention against the endless scroll. A wall of text sitting static for four seconds gives the eye nothing new to track and makes the video feel slower than it is.
When a sentence runs long, split it at a natural grammatical pause rather than an arbitrary word count cutoff — break after a clause, not mid-phrase. 'This cream changed my skin in a week' should break as 'This cream changed my skin' then 'in a week,' not 'This cream changed' then 'my skin in a week,' which reads awkwardly even for a fraction of a second.
Pro tip:Read your caption timing back at normal speed with the sound off. If you can't follow the story from text alone, the captions are either too long per block or not synced tightly enough to the audio.
Takes 3 minutes
Keep captions in the center third of the vertical frame, well clear of the bottom 15 to 20% of the screen where TikTok's UI — caption bar, sound name, engagement buttons — sits, and clear of the right edge where the like, comment and share icons live on most platforms. A caption technically visible in the preview but hidden by a platform's own interface once posted is a wasted caption.
Safe zones differ slightly by platform and by whether a video runs as an organic post or a paid ad, since ad units sometimes add extra UI elements like a 'Shop Now' button that organic posts don't have. When in doubt, keep the bottom quarter and the top tenth of the frame completely clear of any text, and you'll be safe across TikTok, Reels, Shorts and Feed placements without needing a different caption layout for each.
Test placement by exporting the video and viewing it inside the actual platform app rather than trusting the raw export preview, since preview tools rarely simulate the exact UI overlay a live post will have. A five-minute check by posting privately or using a platform's ad preview tool catches placement problems that are invisible in an editing timeline.
Pro tip:When your product or CTA text needs to sit near the bottom of the frame, nudge it up an extra 5% beyond what looks necessary. Platforms occasionally adjust UI element sizing, and a small buffer protects you from future overlap.
Takes 4 minutes
Highlight the specific words in each caption line that carry the core meaning — a number, a benefit, or the word right before a pause — using a different color, a heavier weight, or a subtle scale-up rather than emphasizing everything or nothing. Emphasis only works as a signal when it's used selectively.
Numbers and comparison words are almost always worth emphasizing on their own: '3 days,' 'half the price,' 'twice as fast' catch the eye even in a quick scroll-past glance and often communicate the core claim before a viewer has processed the rest of the sentence. Emphasizing an entire caption line in bright color defeats the purpose, since nothing stands out when everything does.
Match emphasis color to your brand palette rather than defaulting to whatever preset looks flashiest. A consistent accent color used only for emphasized words across all your video ads starts to function as a subtle brand signal over time, the same way a consistent thumbnail style does, without needing a logo on screen.
Takes 3 minutes
Add a semi-transparent background box, a drop shadow, or a text outline behind every caption line rather than relying on font color alone against the raw video. Backgrounds and footage shift between light and dark within the same video, and a caption readable on one frame can disappear entirely two seconds later without some form of consistent contrast treatment.
A thin dark outline (a stroke of two to three pixels) around white text is the most reliable default for footage with mixed or unpredictable lighting, since it guarantees readability against both bright and dark backgrounds without needing a full background box that can feel visually heavy. Reserve solid background boxes for cases where the outline alone still isn't cutting through a genuinely busy shot.
Check contrast specifically during the demo beat, where B-roll often introduces new backgrounds, textures and lighting that differ from the avatar's talking-head shots. It's common for a caption style tested and approved against the hook's clean background to become unreadable the moment the video cuts to a busier product shot.
Takes 2 minutes
Before publishing anything, watch the finished video on a physical phone screen — not a desktop preview — with the sound muted, at the size and distance you'd actually scroll past it. Confirm every caption is fully legible, correctly timed, inside the safe zone, and that emphasis lands on the right words throughout the whole video, not just the hook.
Desktop editing software renders captions larger and sharper than they'll ever appear on a phone screen, which means a caption that looks perfectly fine while editing can be genuinely hard to read once it's actually published and viewed at real scroll size. This gap is one of the most common, and most avoidable, reasons a good script underperforms.
Pro tip:Do this phone check for every new video, not just the first one you style. Caption presets can drift or clip differently against new footage even when the settings haven't technically changed.
Captions aren't decoration on top of a finished video — for the roughly 85% of viewers watching without sound, they are the ad. Bold, correctly sized text, four-to-six-word bursts synced to speech, placement clear of platform UI, selective emphasis on the words that carry meaning, and contrast that survives a busy background will turn a script that reads well into a video that actually communicates on mute. Build these choices into a reusable style once, check it on an actual phone every time, and captions stop being a production afterthought and start being one of your most reliable performance levers.
Minor trims for readability are fine — dropping filler words like 'um' or 'you know' — but the core wording should match closely. Captions that diverge too far from the audio create a subtle disconnect viewers notice even without focusing on it directly.
Aim for captions that occupy roughly 8 to 10% of the frame's height. Test by holding the phone at a normal arm's length and glancing at it for one second — if you have to lean in or concentrate, it's too small.
Not usually. Keeping captions inside the center third of the frame and clear of the bottom 15 to 20% works safely across TikTok, Reels and Shorts. The main difference to check is whether you're running the video as an organic post or a paid ad, since ad units sometimes add extra UI.
Yes, captions are auto-generated and styled for a native social look when you export a video. You can still adjust size, emphasis and placement in the built-in editor before publishing.
Create your first AI UGC video ad in minutes — no filming, no actors, no editing.
Try Klip Kanvas freeFeed ChatGPT real customer language, lock a four-beat UGC structure, force spoken English, and strip the model tells before anyone films or renders the ad.
Use Gemini in Asset Studio, Demand Gen, and Performance Max as a draft engine: feed a real brief, rewrite every line against claims, and keep human UGC as the YouTube-native asset.
Paste a product link and Klip Kanvas writes the script, casts the creator and renders the ad — no filming, no actors, no editing.
