Caption Generator: Add Captions to Any Video (2026)

TL;DR — pick your path

How to add captions to a video (3 steps)

1. Upload — drop your video (MP4, MOV, WebM, MKV, and more) into DeluxeScribe. Files up to 5 GB.
2. Auto-generate — captions are generated automatically with timestamps synced to the video clock. Language auto-detects on the first ~30 seconds; you can override.
3. Export or burn-in — download as SRT or WebVTT (soft captions, viewer toggle) or use the FFmpeg commands below to burn captions into the video permanently.

Caption vs subtitle vs SRT — quick disambiguation

These three words are used interchangeably in casual speech but mean different things in production and accessibility contexts.

Full explainer on the accessibility distinctions: What Is Closed Captioning?

Soft captions vs burned-in — pick the right one

Soft captions (recommended for accessibility)

The caption file is embedded as a separate track inside the video container. The viewer toggles them on or off. Works for YouTube, Vimeo, Premiere delivery, streaming platforms.

ffmpeg -i video.mp4 -i captions.srt -c copy -c:s mov_text output.mp4

-c copy keeps the video and audio streams untouched (no re-encode). -c:s mov_text converts the SRT to the MOV Text format required inside an MP4 container. Instant, no quality loss.

Burned-in captions (required for social)

Captions are permanently rendered into the picture. Required for Instagram, TikTok, Facebook, and any platform that doesn’t support subtitle tracks.

ffmpeg -i video.mp4 -vf subtitles=captions.srt output_burned.mp4

This re-encodes the video (takes longer). For control over font, size, color, position, add force_style — see the styling section below.

Decision table

Styling captions — full force_style reference

FFmpeg’s subtitles filter accepts a force_style parameter that overrides the default caption appearance. Nobody publishes this cheat sheet cleanly, so here it is.

Basic syntax

ffmpeg -i video.mp4 -vf "subtitles=captions.srt:force_style='FontName=Inter,FontSize=24,PrimaryColour=&H00FFFFFF,OutlineColour=&H80000000,BorderStyle=3,Outline=1,Shadow=0,MarginV=40'" output.mp4

force_style parameters

Preset styles for common platforms

Instagram Reel / TikTok (bold, centered, large)

force_style='FontName=Inter,FontSize=32,Bold=1,PrimaryColour=&H00FFFFFF,OutlineColour=&H80000000,BorderStyle=3,MarginV=100,Alignment=2'

YouTube (subtle, standard subtitle style)

force_style='FontName=Roboto,FontSize=22,PrimaryColour=&H00FFFFFF,OutlineColour=&H80000000,BorderStyle=1,Outline=1,MarginV=40'

Cinematic (yellow classic subtitles)

force_style='FontName=Arial,FontSize=24,PrimaryColour=&H0000FFFF,OutlineColour=&H00000000,BorderStyle=1,Outline=2,MarginV=30'

Colour format gotcha: FFmpeg uses &HAABBGGRR where AA is alpha (00 = opaque), BB is blue, GG is green, RR is red — backward from typical web hex codes. Yellow is &H0000FFFF, not &H00FFFF00.

Accuracy by video source

The video container has no effect on caption accuracy — it’s entirely the audio inside it. Common video sources with typical Word Error Rate (WER):

For the full WER breakdown by language and Whisper model size, see How Accurate Is Whisper.

Language support — the honest picture

DeluxeScribe supports 99 languages, but performance varies across them. Same story on every AI captioning tool — nobody achieves equal accuracy in Bengali as in English.

For translated captions (audio in one language, captions in another), transcribe first in the source language, then translate the .srt with DeepL, Google Translate, or an LLM. Watch the length differential — German is ~30% longer than English, French ~15% longer, which breaks caption timing unless you adjust.

Common failures

Caption timing drifts on long videos

Variable frame rate (VFR) videos — common from smartphones — cause timestamps to drift. Convert to constant frame rate before captioning:

ffmpeg -i input.mp4 -vf fps=30 -c:a copy output_cfr.mp4

Wrong language auto-detected

Auto-detection reads the first ~30 seconds. If the video opens with a different language than the main content, set the target language manually.

Missing captions during background music

Heavy music drowns out speech. Source clean audio before mixing, or use vocal isolation (Ultimate Vocal Remover, Krisp) before uploading.

Emoji and special characters break force_style

FFmpeg’s subtitle filter can’t render most emoji or unusual Unicode without a font that includes them. Use an emoji-capable font (Twemoji Mozilla, Noto Color Emoji) or strip emoji from the .srt before burn-in.

Captions too small on vertical video

Default FontSize is calibrated for landscape. For 1080×1920 vertical video, use FontSize=36 or larger and MarginV=100 to keep captions above the platform UI overlay.

Not what you’re looking for?

“Caption generator” also means writing Instagram/LinkedIn post captions— text that goes in the post body, not on the video. That’s a different tool category (LLM writers, hook generators). We don’t do that here.

If you need social-media post copy: try Hootsuite’s AI caption generator, Ahrefs’ Instagram caption generator, or an LLM (ChatGPT, Claude) with a template prompt.

What we do here: generate captions on the video itself — from the audio track — for YouTube, streaming, silent-scroll social, and accessibility compliance.