Your voice, not an echo of Native's

"Platform fingerprint" on "Voice & language" sets how captions sound on a platform. It now learns only from what a person wrote, and needs at least five such posts. Native also looks for lasting wishes in your edits and in what you ask in the Create chat.

What's new

• A caption that began as Native's counts only when enough of its words are new. Cutting or approving it is not enough

• An older fingerprint stays in use until Native next reads your posts, with no set date. "Re-extract" does it now, and replaces your hand edits

• For a company brand, "Brand personality" now sets the feeling of captions. Your own posts still set how they sound

• Taking out all emojis or swapping one word everywhere can become a rule. Most edits, such as a corrected date, price or typo, teach nothing

• Native saves these rules without asking, for new posts only, and marks them as its own: "Added from feedback" or "Learned from your edits"

• In the Create chat, "Remembered for future posts" appears under the reply, with "Undo"

• Native reads up to 30 edits and requests a brand a day, at its own cost