What you need
- Claude Code in a terminal, pointed at a folder with your footage.
- ffmpeg and ffprobe on your PATH.
- A local Whisper model for word-level transcripts.
- Python with numpy, soundfile and OpenCV for the checks.
Captions from word timings
Whisper gives a start and end for every word. Claude groups them into short lines and writes an SRT. After any edit, it rebuilds the timings from the cut, never by shifting the old ones.
Transcribe the final cut
Caption the edited file, not the raw one. Word timestamps on.
Group into lines
A few words per line on 9:16, broken at phrase boundaries.
Write the SRT
One file per deliverable. YouTube and LinkedIn take it as a sidecar.
Burn in for social
For Shorts, Reels and TikTok, burn them into the picture.
Gate the render
Build each caption on its own, scan its pixels for anything in the padding or touching the edge, and refuse to encode on a failure.
ffmpeg -i out/clip_9x16.mp4 -vf "subtitles=out/clip.srt:force_style='FontName=Inter,FontSize=14,Bold=1,Alignment=2,MarginV=90'" out/clip_captioned.mp4Shrink to fit, then look
Never set a fixed font size on variable copy. Measure each line against the space it has and shrink until it fits, then render a still and look at it before any video renders.
A render costs minutes; a caption card costs a second. Sprites builds every card standalone, writes a contact sheet, and the renderer exits non-zero if one fails. Ask Claude for the same: stills first, video second.
ffmpeg -i out/clip.mp4 -vf "fps=1/2,scale=320:-1,tile=6x5" -frames:v 1 qa/sheet.pngEvery Claude video editing guide
Frequently asked questions
How do I auto caption a video with AI for free?
Run a local Whisper model for word timestamps, have Claude Code group them into an SRT, and burn it in with ffmpeg's subtitles filter. All three run on your machine.
How do I add subtitles to a video with AI?
Transcribe the final edit with word timestamps, write an SRT from them, then upload it as a sidecar or burn it into the picture with ffmpeg.
Why are my captions out of sync after editing?
They were shifted from the original transcript instead of rebuilt from the cut. Re-transcribe the edited file, or rebuild word times from the edit list.
Want finished ads, not editing sessions?
Director by Sprites is a full creative team for your startup. Every Monday you get finished, on-brand video ads, reviewed by a human. Strategy, hooks, scripts, shooting and editing, end to end. You never write a prompt. You pick the winners. Sprites ships the campaigns. In private beta.