Around 80% of people watch social video with the sound off. That number, cited in the Verizon Media/Publicis study, hasn't changed meaningfully in years — if anything, the scroll has gotten faster and the tolerance for un-captioned content has gotten lower. Videos with burned-in captions get roughly 40% more watch time than those without, according to Kapwing creator data. But the caption style you pick changes how a clip performs — a bold Hormozi look reads very differently from a minimal editorial fade, and using the wrong one for your platform or your content undermines the lift you'd otherwise get.
Here are 18 burned-in caption styles for shorts, what each one looks like, and when to use it — from the chunky Hormozi preset to a clean cinematic minimal.
Why Burned-In Captions Beat Auto-Captions for Shorts
Auto-captions — the kind platforms generate after upload — are optional to the viewer. They can be turned off, they render in the platform's default font, and they don't sync to the energy of the content. A viewer scrolling past your clip in a feed doesn't have time to enable them.
Burned-in captions are part of the video itself. They're always on. They're styled to match the content's energy. And when they're built from word-level timestamps — as opposed to guessed line breaks — they highlight each word exactly as it's spoken, which creates a reading rhythm that pulls the viewer forward rather than making them parse a wall of text.
Burned-in captions also can't be stripped by the platform. Some platforms, particularly LinkedIn and X, have notoriously unreliable auto-caption coverage. A burned-in clip works identically on every platform, in every embed, in every repurpose.
Word-level timing matters specifically because it's built from real transcription timestamps, not approximated from average speaking speed. Groups break on punctuation or a pause of 0.35 seconds or longer. This means captions don't flicker during natural gaps in speech — a common problem with guessed timing that makes viewers re-read the same line twice.
For a deeper look at how word-level timestamps are generated from the transcription layer, the article on how AI scores podcast clips covers the transcription step in detail.
What Makes a Good Caption Style
Before covering the 18 styles, it helps to understand the dimensions that separate them. A caption style is defined by roughly five things working together.
Readability is the first. Stroke weight (the outline around letters) determines whether text is legible on a light background, a dark background, or both. High-contrast presets like Hormozi and Beast use a heavy black stroke specifically so the text reads on any footage without needing a text box or drop shadow behind it.
Words per line is the second. Three words per line forces the viewer's eye to stay near the bottom of the frame, with minimal horizontal movement. Five words per line covers more of the transcript per screen update but requires slightly more reading effort. For high-energy short-form, fewer words per line generally performs better.
Anchor position is the third. Bottom anchor is standard — it leaves the subject's face clear and is where viewers expect to read. Centre anchor works for cinematic or podcast-style clips where the composition is symmetrical. Top anchor is unusual but effective for vertical clips where the action happens at the bottom of the frame.
Spoken word highlight is the fourth. Some styles highlight the word being spoken in a contrasting color — yellow, cyan, green — in real time. This is the karaoke mechanic, and it works because it gives the viewer a reading cursor: they always know where in the sentence the audio is. Styles without highlighting show the whole group at once.
Animation is the fifth. Bounce, pop, fade, and motion presets add kinetic energy between caption groups. They increase visual interest but can be distracting in clips that need the viewer focused on complex information. Match animation intensity to content energy.
The 18 Caption Styles, and When to Use Each
Bold and Punchy — for High-Energy Hooks
This family is designed for stops-the-scroll performance on TikTok and Reels. Heavy stroke weight, uppercase or semi-uppercase, 3 words per line, bottom anchor.
Hormozi is the reference preset in this category. Chunky uppercase white letters, heavy black stroke, yellow spoken word highlight, 3 words per line, bottom anchor. The yellow highlight against the white text is immediately readable on any footage. Use it for any clip where the hook is a strong claim or a bold statement — the style telegraphs authority before the viewer has processed a single word.
Beast (MrBeast-style) runs extra-bold white with a thick stroke and a green spoken word pop. It carries 4 words per line rather than 3, which gives it slightly more coverage while keeping the punchy feel. Use it for challenge-style or personality-driven content where the energy is high throughout, not just in the hook.
Bounce adds a per-word bounce animation on top of a standard bold white style. The motion is subtle — a small vertical pop per word — but it adds rhythm that matches fast-talking delivery especially well. Use it when the speaker's cadence is quick and you want the captions to mirror that energy rather than sitting static.
Punchline is designed for setups and payoffs. The style holds relatively flat until the punchline word, which gets a size or color pop. Use it for comedy clips, reveals, or any content that has a clear structural beat the viewer should register.
Clean and Editorial — for a Premium Look
This family is for content that should read as thoughtful, polished, or premium. Thinner strokes, sentence case, typically 4-5 words per line, minimal animation.
Minimal is white sentence-case text with a thin stroke or subtle drop shadow. No highlight color, no animation, 5 words per line. It reads neutrally on almost any footage and doesn't compete with the content for attention. Use it for LinkedIn, long-quote clips, educational content where the spoken words are the point and you don't want visual noise.
Negative inverts the color logic: black or dark text on a light semi-transparent pill or box. It reads very differently from white-on-dark and signals a design-forward, editorial sensibility. Use it for lifestyle, beauty, or consumer brand content where the overall visual palette is light.
Cinematic Cut pairs a clean sans-serif with a letterbox crop aesthetic. The captions sit in the black bar beneath the cinematic frame rather than overlaying the footage. Use it for narrative-driven clips, mini-documentaries, or clips that are being repurposed from longer-form interview content where the gravitas should be preserved.
Soft Focus uses a slight blur or fade on the non-active words, keeping the current spoken group sharp. It creates an attention focal point without any animation or color. Use it for meditative, reflective, or emotional content where the pacing is slower and the viewer should sit with each thought.
Colorful and Animated — for Attention
This family maximizes visual engagement. It's appropriate for high-energy platforms and younger audiences, and it should be used with intention — the style competes with the content for attention, so the content needs to be strong enough to hold up.
Neon places the text against a neon glow — the stroke or shadow glows in a saturated color rather than being a hard outline. The effect reads well on dark footage and is immediately distinctive. Use it for music clips, late-night-style content, or anything where the visual aesthetic should feel electric.
Motion adds kinetic entrance and exit transitions to each caption group. Words or lines fly in from one side and exit to the other. Use it for product content, fast-cut edits, or anywhere that the visual pacing of the edit is already high and you want the captions to match that rhythm.
Ember uses a warm orange-to-yellow gradient treatment on the text, as if the words are lit from below. It reads as energetic but more premium than the pure neon treatments. Use it for personal brand content, motivational clips, or any footage with warm color grading.
Highlighter draws a colored highlight behind the current spoken word, mimicking a physical highlighter mark. Use it for educational or explainer content where you want to draw the viewer's attention to the key term in each sentence.
Headline uses a larger font size and tighter line spacing than standard presets. Each group reads like a newspaper headline. Use it for announcement-style clips, launches, or anything where you want the caption to feel declarative rather than conversational.
Karaoke-Style Fills — for Rhythm
These styles track delivery rhythm by filling or coloring the text progressively as the audio moves through each word.
Karaoke is the classic fill: sentence-case text starts in a muted white and fills to a bright cyan as each word is spoken. 5 words per line, bottom anchor. It's the most literal reading-cursor style and works best when the speaker has a rhythmic, even delivery that creates a satisfying fill pattern. Use it for songs, spoken word, or podcast clips where the speaker has a distinctive cadence.
Typewriter reveals each word one character at a time, giving the appearance of live typing. The timing is driven by the real word timestamps rather than a fixed character-per-second rate, so it stays in sync with the audio even for fast or slow delivery. Use it for quotes, reflective voiceover, or any content where the mood is intimate or suspenseful.
Centered and Cinematic — for Podcasts and Long Quotes
This family is optimized for vertical talking-head footage and long quotes where the caption needs to frame the speaker's thoughts rather than compete with them.
Podcast places the captions in the lower-center of the frame with a semi-transparent dark pill background. 4-5 words per line, centered alignment. It's the most readable style for rapid back-and-forth conversation and is immediately recognizable as the podcast format that viewers have been conditioned to follow. Use it for podcast clips, interview content, or any two-person conversation.
Cinetop breaks the bottom-anchor convention by placing captions at the top center of the frame. This works specifically for vertical clips where the speaker's face occupies the lower half of the frame and a bottom caption would overlay their mouth — the most expressive part of the image. Use it for close-up face-to-camera clips where the speaking person fills the bottom third of the frame.
Boxed encloses the full caption group in a solid or semi-transparent box. The box provides guaranteed contrast regardless of the footage behind it, making it the most accessible choice for mixed backgrounds. Use it when the footage has high movement or complex backgrounds where stroke weight alone doesn't guarantee readability.
How to Pick a Caption Style for Your Platform
TikTok rewards visual energy. Bold, animated, or karaoke-style presets perform well because they match the platform's native energy and the audience's trained expectation. Hormozi and Beast are safe defaults. Bounce and Motion work well for fast-cut content.
Instagram Reels skews slightly older and more design-conscious than TikTok. Clean presets like Minimal and Negative are more appropriate for lifestyle and brand content. Bold presets still work for high-energy clips.
YouTube Shorts is the most content-agnostic of the three. Because Shorts appear between regular YouTube videos, the audience arrives with a slightly longer attention span. Podcast and Cinematic Cut perform well here for educational or interview content.
LinkedIn expects professionalism. Minimal and Negative are the right choices. Avoid animated or neon presets — they read as out of place and lower the perceived credibility of the content.
Why Word-Level Timing Matters
Platform auto-captions guess timing from average speaking speed and apply it uniformly. This produces the characteristic subtitle lag — the text appears slightly after or before the word, and line breaks fall mid-thought.
Burned-in captions built from Whisper word timestamps are different. Each word has its own timestamp from the actual transcription of the audio. Groups break on punctuation or a 0.35-second pause in speech. The result is caption groups that match how the speaker actually phrases their thoughts — which means the viewer's eye reads the caption at the same moment the word arrives in the audio, reinforcing both comprehension and retention.
It also means captions don't flicker during natural pauses. When a speaker pauses for emphasis, the previous caption group stays on screen rather than disappearing and reappearing with a partial new group. That stability is what makes word-level captions feel calm and readable rather than jittery.
Preview all 18 styles on the live site to see how each preset looks on real footage before choosing.
Try the AI caption generator with your own video — 10 free credits, no card.
Caption Style Honesty Check
A few things to set realistic expectations.
The style previews shown in the tool are a CSS approximation of the final rendered output. The actual burned-in render uses libass, a subtitle rendering library. For most Latin presets, the visual difference between preview and render is minimal. For non-Latin scripts (Arabic, Devanagari, CJK characters), the display font substitutes to a system font because the branded presets use Latin display typefaces. The rendering is functional but won't match the branded style.
Anchor options are bottom, center, and top. Other anchor positions (left third, right third) are not currently supported.
Font size scales with aspect ratio. The presets are calibrated for 9:16 vertical at 1080x1920. If you export at a lower resolution or a different aspect ratio, the visual weight of the captions will shift slightly.
Frequently Asked Questions
What are burned-in captions? Burned-in captions are text rendered directly into the video pixels during export, making them a permanent part of the video file. Unlike platform auto-captions (which are a separate overlay the platform generates), burned-in captions are always visible and cannot be turned off by the viewer.
Are burned-in captions better than auto-generated captions? For short-form social video, yes — for three reasons. They're always on with no viewer action required. They can be styled to match the content's energy. And they survive repurposing across platforms, embeds, and DMs where platform captions don't travel.
Which caption style is best for TikTok / Reels / Shorts? For TikTok and Reels: Hormozi, Beast, Bounce, or Karaoke for high-energy content; Minimal or Podcast for informational or interview content. For YouTube Shorts: Podcast or Cinematic Cut for longer-form clips; Hormozi or Beast for punchy moments. For LinkedIn: Minimal or Negative only.
What are word-by-word (word-level) captions? Word-by-word captions highlight or reveal each word exactly as it's spoken, using timestamps from transcription rather than approximated timing. The result is captions that stay in sync with delivery and create a reading rhythm that holds viewer attention.
Can I change the caption style after generating clips? Yes. Caption style is selected at export time, not at clip generation time. You can generate a clip once and export it in multiple styles to test which performs better on a given platform.
Do burned-in captions work in languages other than English? The word-level transcription and timing work for any language Whisper supports. The preset display fonts (Hormozi, Beast, and similar bold styles) use Latin display typefaces, so non-Latin scripts (Arabic, Hindi, Chinese, Japanese, Korean) will substitute to a system font. The timing and grouping are accurate; the branded visual style doesn't fully transfer.