Auto-generate captions and edit

Auto‑generate English captions for the May 2 tutorial, then edit them to correct technical terms like ‘SMPTE’, ‘LTC’, and the speaker’s name Kaito, and produce a clean downloadable SRT and closed‑caption track. Set visibility to unlisted for QA and deliver by tomorrow morning.

Before I hand the auto‑generated captions to the content lead, make them easy to approve: surface all low‑confidence words, flag repeated mis‑transcriptions of technical terms and names, suggest correct spellings for the top five flagged words, and indicate where manual timing tweaks are necessary for readability.

I ran auto captions on the May 2 tutorial and the transcript repeatedly misrenders the product name and speaker Kaito’s surname. The host will refuse to sign off if names are wrong. I don’t know whether to edit manually or retrain the model with a custom word list. What’s the likely cause and the fastest reliable next move so I can present an approved caption draft to the host this afternoon?

Every technical video needs significant manual correction after auto‑captions because of jargon, names, and confidence errors. It costs editors hours and slows publishing. What habit change and one recurring small investment (tooling, glossary, or template) will cut average caption cleanup time by at least 50 percent over the next three months?