How to add subtitles to a lecture video automatically
2026-10-20 · 6 min read
Three routes: editor auto-captions (saves typing, not proofing), screen recording plus post (highest rework), or subtitles generated from the narration script — no recognition errors, and only the edited page is redone. If the deck already exists, the third has the lowest total cost.
A 15-minute microlecture usually costs one to two hours of subtitle work: play a line, pause, type it, drag the timing. The real pain arrives when the script changes — one edited sentence and the whole block has to be re-timed.
Subtitles shouldn't be a post-production chore. If the video is generated from a narration script, the subtitles can grow from that same script. Here's the maths on the three common routes.
Why subtitles stopped being optional
Three reasons, none of them about looks:
- Completion rate: most knowledge videos are opened muted, so no subtitles means no content.
- Review and compliance: course platforms require captions, and microlecture competition rubrics score them.
- Discovery: subtitle text is how a platform understands what you actually taught — titles and tags can't guess your key points.
Route 1: auto-captioning inside an editor
CapCut, Premiere and most editors ship a "recognise subtitles" button. Under the hood it's speech-to-text (ASR): record first, then let a model turn audio into text.
It saves you typing, not proofreading or alignment. Domain vocabulary is where it fails — "Fourier transform" becomes "four-year transform", student names and acronyms get mangled. Fixing line by line often costs as much as typing from scratch.
Flubs, pauses and filler words are transcribed verbatim too, so you still have to go back and cut the audio.
Route 2: screen recording plus post captions
Record the deck in Keynote or PowerPoint, then drop the video into an editor and add captions. The most familiar workflow, and the most concentrated set of problems: two apps, two export passes, and sync fixed by hand.
The real cost isn't the software licence, it's rework — change one slide and you re-record, re-time and re-export. Two or three hours per video, ten-plus hours for a week of three.
Route 3: generate subtitles from the narration script
The third route reverses the order. The narration script comes first; the voiceover and the subtitles are both generated from that same script. Subtitles aren't transcribed, they're authored — aligned sentence by sentence, with no recognition errors at all.
In SlideTalk, AI drafts the narration page by page after you import your deck. Edit the script until it's right, then generate the voice: subtitles are split at semantic boundaries and burned into the frame, with no mid-word breaks.
Five caption style presets (including the big-caption short-video look) come with a sidecar .srt file you can re-edit or upload to a course platform.
Changing one sentence only regenerates that page's voice and captions. Everything downstream of the old timing stays put — the thing manual alignment can never do.
Which route fits
Choose by the material you already have, not by brand recognition:
- You have a finished video and only need captions: editor auto-captions, least effort.
- You're still screen-recording and publish occasionally: recording plus post is fine.
- The deck exists, output must be steady, and it will change: generate subtitles from the script — lowest rework cost.
- You need bilingual subtitles: specify English or bilingual when AI writes the script; captions follow the voice instead of being translated and re-timed later.
Cost and common questions
Subtitles aren't billed separately: the script is 1 credit/page and the voice 1 credit/10 characters, and captions come with the voice. A 10-page microlecture costs about 352 credits (≈$0.58), and 600 welcome credits cover your first run. Export, rendering and script edits are free — see pricing.
For the full educator workflow see the educator use case, and for the end-to-end route see how to turn a PPT into a microlecture.