WhatsApp folds Meta AI into chat themes and video call backgrounds

What actually shipped in this batch

Meta grouped several months of WhatsApp changes into one post. The user-facing list: Live Photos on iOS and Motion Photos on Android, both adding sound and movement to shared captures; AI-generated chat themes; AI-generated video call backgrounds that also work when taking photos and videos in chats; two new sticker packs (Fearless Bird and Vacation); group search by member; and document scanning on Android.

Three of these features route through Meta AI: custom chat themes, video call backgrounds, and in-chat photo/video backgrounds. The rest are conventional client features. Meta notes plainly that the AI features may not reach everyone.

You can now use the power of Meta AI to boost your creativity and make custom chat themes of your own. (Meta AI features may not be available to all users.)Montana Labs

The AI is placed at the edit surface, not a separate chatbot

What stands out from a frontend perspective is placement. Meta did not add another way to talk to an assistant. It wired generation into the moments where a user is already choosing a look — the theme picker, the video call background control, the in-chat camera. The generation output is a customization asset, not a conversation.

That framing matters because it sets a low bar for correctness. A generated chat theme or call background has no wrong answer the way a factual reply does. The feature can ship to a subset of users, produce imperfect results, and still be a net positive to the interface. That tolerance is likely why these are the AI surfaces Meta chose to expose first inside a private-messaging app.

Closing the iOS/Android gap, one feature at a time

Two items in this list are pure parity work. Live Photos ship to iOS while Motion Photos ship to Android — the same idea implemented against each platform's native motion-photo format. Document scanning, previously iPhone-only, now lands on Android.

These are not glamorous, but they are the recurring tax of running one product across two mobile platforms with different media APIs. The animated-capture feature in particular has to bridge Apple's Live Photo container and Android's motion photo container into a single shareable object that renders on both. Announcing them alongside the AI features is a reminder that most of the engineering effort in a mature messaging client is still cross-platform plumbing, not model integration.

Group search shows a different, quieter design instinct

The group-search change is the most telling non-AI feature. Instead of asking users to remember a creatively named group, Meta lets you search for a person you know and surfaces the groups you share. It solves a real recall problem by pivoting the query from the hard-to-remember object (the group name) to the easy-to-remember one (a contact).

That is a deterministic, index-driven fix for a problem an LLM would be a poor and expensive tool for. The contrast inside a single announcement is instructive: Meta reached for generation where the output is decorative and low-stakes, and reached for conventional data structures where the user needs a reliable, correct answer.

The implication: Meta is testing AI adoption through cosmetics

By routing its first WhatsApp AI features through themes, call backgrounds, and camera effects, Meta is introducing generative tooling in the lowest-risk possible slot inside an app whose core promise is private conversation. There is no assistant reading messages here — only optional visual generation gated behind an availability flag.

For teams building AI into existing products, the pattern is worth noting: pick surfaces where output quality is subjective, failure is cheap, and the feature enhances something the user was already doing. That lets you gather real usage without staking the product's trust on model reliability. Meta closed the post with 'stay tuned,' which suggests these cosmetic entry points are a deliberate first step rather than the destination.