Words
Do Reeloya captions still need a person?
Yes. The caption draft comes from speech recognition. It misses homophones, names, and numbers. Read every line before anyone sees it. Styled captions, motion, and packaging belong in your finishing editor.
Updated
What is in the draft?
Words from the recording, attached to the footage. The draft is text waiting for a check. It is not a title, a cover, or an animated caption look. Reeloya treats it that way on purpose. A wrong product name in a burned-in caption is a published error, and the model does not know which spelling you sell.
What do you check?
- Names, brands, model numbers, and amounts, against the original recording.
- Homophones, especially product names.
- Promises such as “always,” “forever,” or “free.” Confirm the speaker said that word.
The same list is in how to turn a long video into short videos.
Why not ship the first transcription?
Speech recognition is a useful first pass. It is weak where the word is rare, where two words sound the same, and where a number matters. Glancing at a preview is not the check. Read the line.
Where do styled captions go?
Font, position, color, and motion are finishing. Take the kept lines into the editor you already use for packaging. Apps that advertise automatic captions often mean a styled track inside their own export. That is a different job from a proofreading draft. The split is also in how Reeloya differs from CutAI.