Adding subtitles is one of the highest-return edits you can make. Captions increase watch time by roughly 12% to 40%, and they let people follow your video in the sound-off scroll where most of TikTok, Reels, and Shorts is actually watched. Around 80% of viral short-form clips use burned-in captions, and most animate them word by word.

But "add subtitles" hides three genuinely different approaches, and picking the wrong one costs you either time or reach. Here are all three, plus the vocabulary, the platform rules, and the tools.

First, the vocabulary: subtitles vs. captions vs. SRT

These three words get used interchangeably, but they mean different things.

  • Subtitles assume the viewer can hear the audio. They transcribe or translate spoken dialogue only.
  • Captions (closed captions) assume the viewer cannot hear. They include dialogue plus sound effects, music cues, and speaker labels, and viewers can toggle them on or off.
  • SRT is a file format, not a type of subtitle. An .srt file is a plain-text list of caption lines with start and end timecodes. It is universal and works everywhere, especially YouTube. VTT is the closely related web standard for HTML5 players.

So SRT is how subtitles or captions are stored and moved between tools; subtitles and captions describe what the text contains.

A video frame with captions burned in, one word highlighted as it is spoken.

The three ways to add subtitles

1. Auto-transcription (fastest)

Modern tools listen to your audio and generate timed captions automatically. For clear, standard English, most tools land around 95% to 98% accuracy; platform-native auto-captions (like TikTok's) tend to sit a little lower, around 85% to 90%. Accuracy drops with accents, technical terms, background noise, or multiple speakers, so always proofread.

Auto-transcription is the right default for almost everyone in 2026. You get 90% of the way there in seconds, then fix the handful of wrong words.

How it usually works:

  1. Import or upload your video.
  2. Click the auto-subtitle or auto-caption button and pick the spoken language.
  3. Review the generated text and correct names, jargon, and misheard words.
  4. Restyle, reposition, and either burn in or export a file.

2. Importing or writing an SRT (most accurate, toggleable)

If you want subtitles viewers can turn on and off - or you need broadcast-grade accuracy - use an SRT (or VTT) file. You can generate one automatically and then hand-correct it, or write it from scratch. Each entry has a number, a start-to-end timecode, and the text.

Upload the SRT alongside your video on YouTube, Vimeo, or a website player, and the platform overlays it while letting viewers control visibility. This is the accessibility-friendly, multi-language-friendly path, and it keeps your text editable forever. The downside is that it is the most time-consuming to perfect by hand.

3. Burning captions into the video (best for social feeds)

Burned-in (or "hardcoded") captions are drawn permanently onto the video frames. They always show, no matter where the file is played, re-uploaded, or re-shared, and you have full control over font, color, size, and animation.

This is the winning approach for TikTok, Instagram Reels, and YouTube Shorts, where videos autoplay muted in the feed and you cannot rely on the viewer to switch captions on. The trade-off: burned-in text cannot be turned off or edited after export, so proofread before you render.

Quick comparison

Method Speed Accuracy Editable later Best use
Auto-transcription Fastest High (proofread) Yes, before export Nearly everyone, as a starting point
SRT file Slowest by hand Highest Yes, anytime YouTube, accessibility, multi-language
Burned-in Fast with auto High (proofread) No, once rendered TikTok, Reels, Shorts, feeds

Note that these combine: the common 2026 workflow is auto-transcribe, correct, then burn in for social while also exporting an SRT for platforms that support it.

Platform tips

  • TikTok: burned-in animated captions dominate viral clips. Keep chunks short, 3 to 7 words, on screen 1 to 3 seconds each, breaking at natural pauses.
  • YouTube: the algorithm rewards watch time, and YouTube reads uploaded SRT files well for search and accessibility. Long-form videos benefit from an SRT; Shorts behave like TikTok, so burn captions in.
  • Instagram Reels: viewers scroll with sound off, so burned-in captions are especially valuable. Instagram optimizes for saves and shares, so make captions readable at a glance.

General formatting rule of thumb: aim for 4 to 6 words per line, split long sentences across blocks, and keep the text clear of platform UI at the bottom and sides.

Tools for adding subtitles

  • Browser tools like VEED, Kapwing, and Happy Scribe auto-generate captions and export SRT/VTT quickly.
  • Premiere Pro, DaVinci Resolve, and CapCut all have built-in speech-to-text captioning for editors already in those apps.
  • Zella generates captions entirely on your device, on Mac and iPhone.

Most cloud tools upload your video to their servers to transcribe it, which is fine for many people but a concern if your footage is sensitive or unreleased.

Zella: on-device auto-captions on Mac and iPhone

Zella takes the fast auto-transcription path without the upload. It is native to Apple devices - Mac (Apple Silicon and Intel, macOS 14 or later), iPhone - and runs 100% on-device, so your video is never sent anywhere to be transcribed. It works in airplane mode.

What you get:

  • Local transcription with word-level timing, so a caption lands on the exact word rather than the rough second.
  • 9 caption presets, including Word Pop, Hormozi, Karaoke, Neon Glow, Clean Minimal, and Beast.
  • Active-word highlight that pops each word as it is spoken - the animated style that performs best on short-form.
  • A transcript word editor and SRT import, so you can fix a misheard name or bring in captions you already have.
  • Custom font import for brand type, in any language.

On the Mac, the Captions tab also exports SRT, VTT or CSV when you need a caption file for YouTube or a web player. On iPhone, which is on the App Store, you get SRT import plus something the Mac does not have: on-device caption translation into 19 languages on iOS 18 or later. See captions and titles on iPhone.

The free plan includes captions and 1080p export up to 2 minutes, watermark-free and with no account. Pro is a single one-time purchase of $89, not a subscription. See how it works on the captions page, or follow the step-by-step guide to auto-caption a video on Mac. Get Zella to caption your first clip.

Frequently asked questions

What is the difference between subtitles and captions?

Subtitles transcribe spoken dialogue for viewers who can hear the audio. Captions add non-speech information like sound effects and speaker labels for viewers who cannot hear, and are toggleable. In casual use people say "captions" for both.

Should I burn subtitles in or upload an SRT?

Burn them in for TikTok, Reels, and Shorts, where videos play muted in the feed. Upload an SRT for YouTube long-form, websites, and accessibility, so viewers can toggle them and search engines can read them. Many creators do both.

How accurate is auto-captioning in 2026?

For clear, standard English, dedicated tools reach roughly 95% to 98%. Platform-native auto-captions are usually lower, around 85% to 90%. Always proofread names, technical terms, and anything spoken with an accent or over noise.

Can I add subtitles without uploading my video to a server?

Yes. On-device tools like Zella (Mac and iPhone) transcribe locally, so your footage never leaves your device. Most browser-based tools upload your file to transcribe it.

What format should I export subtitles in?

SRT is the safest universal choice and works with YouTube. VTT is the web standard for HTML5 players, and CSV opens the transcript in a spreadsheet. Zella's Mac app exports all three from the Captions tab.

Sources: