YouTube generates captions for every Short you upload, but they appear as a closed-caption layer most viewers never enable. If your goal is silent-feed retention, you need captions burned into the pixels — not a toggle the algorithm hides under a chevron.
Option 1 — YouTube's native auto-captions
After upload, YouTube transcribes the audio (usually within minutes for Shorts under 60s) and exposes the captions in the Subtitles tab of YouTube Studio. You can edit them, but the on-feed default is still 'off' for most viewers. Useful as accessibility fallback; not enough as a retention tool.
- Pro: free, accurate enough on clean English, accessibility-compliant.
- Pro: editable inside Studio with the timing already done.
- Con: viewers must opt in — most don't.
- Con: visual style is YouTube's default white sans, no brand control.
Option 2 — Burn captions into the video before upload
Run the clip through an AI captions tool (SoCaptions, CapCut, Submagic) that bakes a styled caption layer into the export. The captions are now part of the pixels — they show up the instant the Short auto-plays in the feed, and they survive cross-posts to TikTok and Reels.
Best for short-form because the silent-scroll moment is the entire battle. If the first 1.5 seconds of your Short don't show readable text, the algorithm sees a swipe instead of a watch.
Option 3 — Upload your own SRT
If you've transcribed elsewhere (Whisper, Rev, Premiere) you can upload the SRT directly in Studio. Same caveat as native auto-captions: closed-only, viewer must enable. Use this when you want accessibility-compliant captions and you're shipping the burned version separately.
What we'd actually do
Burn captions into the export, upload the captioned MP4, and let YouTube generate its accessibility track on top. You get the silent-scroll retention from the burned captions and the WCAG-compliance closed-caption track YouTube provides for free.