By PayPerVideo Editorial Team, October 22, 2026

AI Video Text Overlay Guide: Titles That Get Read

Most AI video is watched on a phone, often with the sound off, inside a feed that covers the edges of the screen with buttons and captions. In that setting, the text you place on a clip is not decoration. It is the part of the video most likely to be read, and the part most likely to go unread if you treat it as an afterthought.

Quick answer: Give each line of overlay text one job, keep it short enough to read at a glance, place it inside the platform's safe zone, and set it against a background with real contrast. Add text in an editor after generating the footage rather than asking the model to render words, and check the result on a phone before publishing.

This guide covers planning, writing, positioning, and checking text overlays for AI-generated clips. It does not verify text or caption tools inside any generator, and it does not claim that overlays raise views or conversions. It covers making the words readable and accurate.

Give each line one job

Text on a short clip usually does one of three things: it hooks attention, it adds context the visuals cannot, or it tells the viewer what to do next. A line that tries to do all three reads as noise.

TikTok's creative best practices advise introducing the content proposition in the first three seconds and prioritizing the hook in the first six, and they recommend captions or text overlays to provide context. On a five or ten second generated clip, that usually means one short line up front and at most one more near the end. If a clip needs three sentences of explanation, the script or the shot is wrong, not the font.

Write the lines before you generate, in the same pass as the shot plan. The AI video storyboard guide covers planning each shot, and the overlay line belongs in that plan, next to the shot it sits on.

Write for reading speed

Viewers read text on video under time pressure. TikTok's best practices recommend displaying five to ten words per second when using text, and that range is a useful ceiling anywhere: a line that stays up for two seconds should hold well under twenty words, and usually far fewer. Short lines win. "Price drop Friday" reads instantly; a full sentence with a subordinate clause does not.

Match the amount of text to the clip's real length. PayPerVideo offers duration tiers of 5, 10, and 15 seconds, with supported duration, resolution, and audio varying by model, so check the actual length of your export before timing lines to it. A five-second clip carries one idea. A fifteen-second clip can carry a short sequence, but each line still needs its own dwell time.

Make the words legible on any background

Generated footage moves. Skies shift, crowds walk, lights flicker, and any of it can pass behind your text mid-sentence. White text that reads perfectly over a dark doorway disappears when the camera pans to a window.

The Web Content Accessibility Guidelines' contrast criterion calls for a contrast ratio of at least 4.5:1 between normal text and its background, and 3:1 for large-scale text. That standard was written for web pages rather than video, but it is the right floor for overlays. The reliable way to hold it over moving footage is to separate the text from the picture: a solid or semi-transparent backing bar, a scrim that darkens the frame behind the words, or a position over a calm area of the scene. Outlined or shadowed text helps less than people expect over busy generated motion.

Keep type large and lines short. Two lines of big type beat four lines of small type on a phone screen.

Stay inside the safe zone

Every full-screen player draws its own interface over your video: usernames, action buttons, progress bars, captions, and calls to action. TikTok's in-feed specification notes that the safe zone depends on the video dimension, the ad caption length, and any added formats, and it provides safe-zone files for checking layouts. Treat the top, bottom, and side edges as occupied territory and keep every word in the clear central band.

This is the same discipline as framing for the right shape, and the AI video aspect ratio guide covers choosing the frame itself. The only trustworthy check is watching the clip in the real player on a phone, with the interface visible. A desktop preview hides the exact elements that cover your text in the feed.

Add text in an editor, not in the prompt

Current video models are unreliable at rendering written words. A prompt that asks for a sign, a title card, or on-screen text can return misspelled, warped, or half-formed lettering, and a failed render spends generation budget. The dependable workflow is to generate a clean scene and add the words afterward in an editor, where you control the exact wording, font, timing, and placement.

This guide does not verify text-overlay or caption features inside PayPerVideo; check the actual options in the AI video generator for your chosen settings, and use an external editor when precision matters. If a concept genuinely needs rendered text inside the scene, such as writing on a storefront, keep it short, inspect every letter in the export, and budget for retries. The AI video artifacts guide covers inspecting generated clips for defects.

Time text to the clip

A line should stay on screen long enough to read twice at a comfortable pace, and it should leave before it overstays. On a five-second clip, one line held for most of the runtime usually beats two lines rushed through. On longer clips, change lines on visual beats rather than mid-gesture, so the cut feels intentional.

If the video has sound, decide whether the text repeats the voice or adds to it, and keep the two from competing. Words the viewer must not miss belong on screen even with sound on, because feeds are often watched muted.

Review before publishing

1. Job: Each line does one thing: hook, context, or call to action.
2. Length: Every line is readable at a glance, within roughly five to ten words per second of screen time.
3. Contrast: Text holds a strong contrast ratio against the footage for its full duration, with a backing bar or scrim where needed.
4. Position: All text sits inside the destination's safe zone, verified in the real player on a phone.
5. Accuracy: Every word is spelled correctly and says something true about the product or offer.

Frequently asked questions

How many words should a video text overlay have?

Fewer than you think. TikTok recommends displaying five to ten words per second when using text, so a line visible for two seconds should carry well under twenty words. Short lines of large type read best on phones.

Should I put text in the prompt or add it afterward?

Add it afterward in an editor. Current video models often render garbled or misspelled lettering, and a failed render spends generation budget. Generate a clean scene, then place exact words where you control wording, timing, and position.

What contrast does overlay text need?

The WCAG contrast criterion calls for at least 4.5:1 between normal text and its background, and 3:1 for large-scale text. Over moving generated footage, hold that floor with a solid backing bar or a darkened scrim behind the words.

Where on the screen should overlay text go?

In the clear central area, away from the top, bottom, and side edges where players draw usernames, buttons, and captions. Verify placement in the actual app on a phone, since each platform's interface differs.

Does PayPerVideo add text overlays or captions automatically?

This guide does not verify those features. Check the options shown for your chosen settings before relying on them, and use an external editor when exact wording and placement matter.

Next step: Write your overlay lines into the shot plan, generate the clean scene in the AI video generator, add the words in an editor, and watch the finished clip in the real player before publishing. Review current pricing for the duration and model options your clip needs.