What should StreetVoca optimize first: OCR speed or word review flow?

StreetVoca is built for language learners who find new words in real photos: signs, menus, books, screenshots, and daily scenes. The current loop is: capture or import a photo, detect candidate words, save useful ones to Wordbook, then review them later.

For early users, which part would most affect whether you keep using it: faster OCR, cleaner word selection, better review reminders, or clearer Free/Plus limits?