Captions look simple from the outside: words on a screen, timed to speech. In practice, a good captioned video is the result of four separate decisions — what the words are, when they appear, how they are grouped, and how they look. Get any one of them wrong and the caption stops helping the viewer. This guide walks through the whole workflow so you can produce clean, readable captions regardless of which tool you use.
1. Start with an accurate transcript
Everything downstream depends on the transcript. Automatic speech recognition (ASR) has improved dramatically, but no model is perfect, especially with names, jargon, accents, overlapping speech, or background music. Treat the first machine transcript as a draft, not a finished product. Read it against the audio once, start to finish, and fix the words before you touch timing. Fixing text after captions are split into timed lines is far more tedious.
- Correct proper nouns, brand names, and technical terms first — these are what ASR gets wrong most often and what viewers notice most.
- Expand or normalise numbers the way you want them read: "twenty twenty-six" vs "2026" is a stylistic choice, but be consistent.
- Decide how to handle filler words ("um", "you know", "like"). Removing them makes captions cleaner but can change meaning or tone — keep them for interviews, cut them for polished promos.
2. Time the captions to speech, not to the clock
A caption should appear when the words are spoken and disappear when they stop being useful. The common failure is captions that lag the audio by a fraction of a second — the brain notices even a 150–200ms mismatch. If your captions feel "off" but you cannot say why, it is almost always a global timing offset. Most editors let you nudge every caption forward or backward by a fixed amount; use that before you hand-adjust individual lines.
Rule of thumb
Aim for a reading speed of roughly 12–17 characters per second. Faster than that and viewers cannot finish the line before it changes; much slower and the caption lingers awkwardly after the speaker has moved on.
3. Group words into readable lines
This is the step most people skip, and it is the difference between captions that feel professional and captions that feel auto-generated. Do not break lines in the middle of a grammatical unit. Keep articles with their nouns, prepositions with their phrases, and names intact. A good line break falls at a natural pause — where a person would breathe if reading aloud.
- Keep lines to about 32–42 characters for horizontal video, shorter for vertical.
- Prefer one to two lines on screen at a time. Three lines cover too much of the frame on mobile.
- Never split a hyphenated word or a URL across two captions.
4. Style captions for where they will be watched
Short-form social video (TikTok, Reels, Shorts) is watched on small screens, often in bright environments, frequently muted. That context dictates the styling: high contrast, a heavy font weight, a stroke or shadow so text survives over any background, and a safe position that avoids platform UI. A caption that looks elegant on your desktop preview can vanish against a white shirt on a phone. Always preview on the aspect ratio and roughly the size people will actually watch.
- 1Pick a bold, legible sans-serif — thin display fonts disappear at small sizes.
- 2Add an outline (stroke) or drop shadow so text is readable over any footage.
- 3Keep captions inside the safe area: away from the top notch and the bottom row of platform buttons.
- 4Limit animation. Word-by-word highlighting can boost retention, but constant motion is exhausting.
5. Export the right format
You have two fundamentally different outputs. Burned-in (open) captions are rendered permanently into the video pixels — universal, uneditable, and required for most social feeds. Sidecar subtitle files (SRT, VTT) sit alongside the video and can be toggled on or off — the right choice for YouTube long-form, courses, and accessibility compliance. Many creators export both: a burned-in version for social and an SRT for platforms that support real subtitle tracks.
In AutoCap you do all five steps in the browser: generate or import a transcript, correct the text, adjust timing and grouping, apply a style preset, and export either a no-watermark captioned video or an SRT file. Because the editing happens locally, your footage stays on your device while you work.
Common mistakes to avoid
- Trusting the raw ASR output without reading it — the one wrong word is always in the thumbnail frame.
- Captions timed to the clock instead of the voice, producing a subtle lag.
- Lines that are too long, breaking at awkward points.
- Low-contrast text with no stroke, unreadable over bright footage.
- Covering the platform UI so the last word is hidden behind a button.