Latest Update

Google DeepMind's TacticAI predicts football plays 8 seconds early

2026.06.12
What's up, Carat crew here. 😄
Today: Google TacticAI reading football 8 seconds ahead, Magnific splitting an image into editable layers, ChatGPT that now organizes its own memory, and a paper plane turned into a Pixar-grade one-take video.

🔥 Google AI sees the play 8 seconds ahead? TacticAI

ⓒ Google DeepMind
Google DeepMind deployed its football tactics AI, TacticAI, with Brazilian powerhouse Palmeiras for the first time. It reads the movement of all 22 players during a match and predicts how the next 8 seconds will unfold.
Before, it only analyzed set pieces like corner kicks; now it reads the flow of live, open play. Staff can even drag players around on screen to test tactics in advance. Liverpool used this tech the season they won the Premier League.

📌 3 more things today

1️⃣ Magnific splits a finished image into layers

ⓒ @magnific
Magnific unveiled 'Auto Layers,' which automatically breaks one finished image into layers like text, subject, and background. No more rebuilding the whole thing just to fix one typo inside an image.
Pick the part you want to change and edit just that layer. It cuts out the most tedious part of design work. You can also touch up just one part of an image on Carat.

2️⃣ ChatGPT started organizing its own memory

ⓒ OpenAI
ⓒ OpenAI
OpenAI added a new memory feature called 'Dreaming' to ChatGPT. As your chats pile up, it sorts out the important bits into memory on its own in the background. You don't have to say 'remember this' every time.
Your memories also update on their own over time. If you note 'I'm having a birthday party next week,' once the day passes it switches to 'had a birthday party.' It's rolling out to Plus and Pro users in the US first.

3️⃣ A single paper plane became a Pixar-grade video

ⓒ @AngryTomtweets
A one-take video of a paper plane gliding through the sky, with no cuts, is catching eyes. The camera follows the plane so smoothly that people say it looks like a Pixar animation.
It's not stitched from multiple cuts but flows as one continuous shot. The creator says it was made with a single workflow and shared the process too. You can make this kind of cinematic video with a video model on Carat too.
👉 Make it on Carat

🧪 Prompt of the day

A 'Chinese wanghong makeover' photo is trending on social right now. It's a concept with doll-like porcelain-white skin, sharp Chinese-style eye makeup, and Hanfu outfits. It started on Xiaohongshu and went so viral that some people fly to China just to try it. The Carat team made one from a single photo.
Generated with ChatGPT Image 2 on Carat
Generated with ChatGPT Image 2 on Carat
Transform my photo with the following prompt using ChatGPT Image 2.
Prompt: Use the uploaded model image as the identity anchor. Preserve the person's recognizable face shape, eyes, nose, lips, skin tone relationship, hairstyle impression, and natural expression, but visibly transform her into a Shanghai wanghong makeover experience. The makeover must affect the face, not only the clothes and background.

MAKEUP MUST BE OBVIOUS AND CENTRAL: apply dramatic Chinese wanghong doll makeup with very pale porcelain-white base makeup, flawless matte filtered skin, smooth beauty-app texture, soft arched brows, sharp lifted eyeliner, opened inner corners, oversized doe-eye circle lens effect, thick manga-style individual cluster false eyelashes, clearly visible lower lashes, exaggerated plump aegyosal under both eyes, bright shimmery aegyosal highlight, diamond glitter on the inner corners and under-eye area, high-placement peach-pink blush directly under the eyes, strong rosy flush across the cheeks, visible red blush on the nose tip and a soft red tint along the nose bridge, delicate contour but not Western glam, glossy syrupy gradient lips with high-shine glass finish. The face should look like a professionally done wanghong costume-photo makeup look, cute, ethereal, slightly doll-like, and highly retouched while still photographic.

Shanghai wanghong costume-photo experience: Hanfu-inspired influencer styling with flowing sheer chiffon layers, embroidered details, delicate ribbons, pearl/floral hair ornaments, dangling crystal hair jewelry, and optional round fan, oil-paper umbrella, lantern, or flower prop. Outfit color should vary naturally each time within elegant palettes such as ivory, blush pink, pale blue, lavender, mint, champagne, jade green, deep red, or black-gold; do not lock to one exact color. Avoid readable logos, brand names, or text.

Imperial glamour concept. Traditional interior, carved wooden window, palace-style corridor, red lantern glow. Close-up beauty portrait with ornate Hanfu-inspired outfit and hairpiece. Pose with chin slightly turned and one hand near hair ornament. Focus on dramatic costume-photo makeup: doll-white face, manga lashes, large circle lenses, heavy under-eye glitter, rosy cheeks and nose, glossy gradient lips.

Camera texture: realistic iPhone photo or friend-taken smartphone portrait, social-media retouch, soft focus, shallow depth of field, Xiaohongshu-style viral makeover post. No text, no captions, no watermark, no readable logo, no duplicated face, no extra merged people, no distorted hands, no celebrity resemblance, no heavy 3D render.
Find more prompts in the prompt gallery. 👉 See the wanghong prompt
That's it for today. We'll be back tomorrow with fresh AI news. ☺️

Was this helpful? Share with your friends

Share

Get daily AI trend updates