Total reading time about 6 minutes.
Welcome to Visually AI!
🔮AI News this Week
ChatGPT 4o Images for ad-style visuals
People have been using ChatGPT 4o images for a variety of useful ad-style images because the quality is impressive, it has the ability for accurate text rendering, and you can make edits within your chat.
I experimented with several and here’s a fun template you can try yourself:
Prompt: Design a 3D miniature store shaped like a [brand item or logo], with a glowing neon [“slogan” or “product name”, max 3–4 words] sign above the entrance.
Incorporate tiny human figures enjoying the store, soft clay textures, and a [color scheme or palette, e.g., pastel colors, brand-themed tones].
The scene should exude a whimsical and playful atmosphere, viewed isometrically with high detail, capturing the essence of the [brand name] brand.
💡 Tips for Customization
-
[brand item or logo]: Choose a recognizable, iconic product shape (e.g., bottle, snack, sneaker, mascot).
-
[“slogan” or “product name”]: Keep it short, 1–4 words max. Use official slogans or playful nicknames.
-
[color scheme or palette]: Stick to brand colors or creative variants (pastels, neons, warm tones).
-
[brand name]: Repeat the brand’s name here to reinforce its personality.
Here’s an example I used:
Design a 3D miniature store shaped like a giant Welch’s® Fruit Snack, with a glowing neon “Sweet Delights” sign above the entrance. Incorporate tiny human figures enjoying the store, soft clay textures, and a pastel color scheme. The scene should exude a whimsical and playful atmosphere, viewed isometrically with high detail, capturing the essence of Welch’s® brand.
I removed the text at the bottom and expanded the image to 16:9 aspect ratio in Midjourney’s Editor feature:
And finally, generated a video with Kling 1.6 (sound added by MMAudio on Replicate).
Kling video prompt: The whimsical “Sweet Delights” ice cream storefront glows under daylight as animated characters serve customers behind the counter, browse colorful shelves, and laugh at tables with melting treats. The scene transitions to a warm evening with golden lights illuminating the pink cone structure, fireflies dancing around cheerful signage, while the camera slowly zooms out to reveal the joyful candy-colored neighborhood.
Here’s a few examples using other brands, slogans, and color palettes in the template above:
Krea ChatGPT Paint
Prompt image edits visually in ChatGPT with Krea GPT Paint to add or remove objects to images with reference images, notes, or shapes. You can enhance your images with new Topaz upscaler or Krea’s own image enhancers.
Krea’s been shipping a lot of new models and tools, including Restyle video, 3D Objects, and video Lip sync using Hedra for animation.
The new version keeps the main composition of the original, but there are definitely differences in the faces, clothes, and poses. It helps to include a detailed prompt to match your original image with the new elements described to fit into the new result.
Here’s the original Midjourney image next to the edited result from Krea GPT Paint:
📸 AI Snapshots
Now you can download your ChatGPT Deep Research reports as formatted PDFs. I generated a report about video advertising trends in the U.S. over the past three years. Read the report from the example below, here.
Higgsfield AI launched Ad templates that let you choose a template, upload an image, and generate an ad-style video quickly. The feature is available for Pro and Ultimate plans. This is my result using a Flux1.1 image with Higgsfield’s Cloud Monument Ad template:
Luma AI added the Reframe feature in Dream Machine that lets you outpaint and resize your uploaded videos. You can even move the video to different areas of the reframe position for your best results:
🌎 AI Developments All Over The Globe
OpenAI and Microsoft are renegotiating their multibillion-dollar partnership. OpenAI wants to reduce Microsoft’s revenue share from 20% to 10% by 2030, while Microsoft is pushing for extended access to OpenAI’s models as OpenAI restructures into a public benefit corporation.
Windsurf launched SWE-1, SWE-1-lite, and SWE-1-mini—developer-focused AI models that use a “flow awareness” system to integrate better with the software engineering process.
Alibaba revealed a new training method called ZeroSearch, which lets LLMs simulate search results without calling real engines; cutting training costs by 88%.
Function Health acquired MRI scanning startup Ezra to integrate full-body AI-powered MRI scans into its platform, aiming to offer complete health diagnostics.
🚀 My Recent Top AI Tools Picks
-
HeyGen Avatar IV: Turn a single photo and script into lifelike videos with realistic expressions and gestures in seconds.
-
Lovart: AI agent design tool collaborates with you on storyboards, visual ads, and more.
-
TikTok AI Alive: Create immersive videos from static photos directly within TikTok Stories.
-
ChatGPT Codex: Cloud-based software engineering agent that can independently navigate your codebase, implement and test code changes, and proposes pull requests for you to review. Rolling out to Pro, Enterprise, and Teams users now.
-
Higgsfield AI: Now supports mixed motion styles (e.g. levitation + drone), giving users full control over stylized movement from a single image.
-
LTXStudio: LTX now has longer scenes, better resolution, and tighter keyframe control for script-to-video creation.
🖼️ Image Prompts
Here’s a prompt you can use in ChatGPT 4o Image:
Zero-Gravity Kitchen Test:
Depict a zero-gravity space capsule kitchen test for a [Brand food item], with ingredients and branded packaging floating mid-prep. A spoonful of sauce freezes in space. Futuristic color-coded kitchen tools. Scientific, clean, playful.
I used Chipotle Burrito Bowls for this example:
I used this prompt to generate images to test Higgsfield AI’s new “Levitation” Effect:
Prompt: A teenager rises slowly above the ground in a suburban backyard at night, limbs relaxed, eyes closed, soft glow surrounding them, stars barely visible through low clouds, sci-fi horror vibe
And for Higgsfield’s “Eyes In” Effect:
Prompt: An anxious woman stares into a mirror in a dim bathroom, her reflection distorted slightly, shadows forming around her eyes, cinematic moody lighting
🎥 Video Prompts
I typed a short prompt, like “The teenage boy levitates and rises above the neighborhood”, and used the “Enhance” option in Higgsfield:
Enhanced Prompt: In a dimly lit suburban neighborhood under a blanket of swirling clouds and shimmering stars, a teenage boy floats effortlessly above the ground. His feet hover nearly a foot off the earth, dressed in casual jeans and a hoodie that rustles softly in the cool night air. The surrounding houses appear still, their silhouettes framed by leafy trees swaying gently in the breeze. As he begins to rise, there is a palpable stillness, magnified by the mystery of the moment. The camera moves slowly around him, capturing the tranquil motion as his hair gently drifts upward, illuminated by an ethereal light spilling from below. A distant rumble of thunder echoes softly, as if the universe itself holds its breath, creating a sublime atmosphere of wonder and liberation.
Another short prompt: “Camera zooms in to her eye show another dimension”
Enhanced Prompt: In a cramped, dimly lit bathroom, the camera lingers on a young woman, her skin glistening with droplets of sweat as she stands before a fogged mirror. The overhead light casts soft shadows, tracing the contours of her solemn expression. The air is thick with unspoken tension, her curly hair framing a face that embodies vulnerability mixed with resolve.
Suddenly, the camera rushes into her eye, the lens piercing through the pupil. As it zooms deeper, the world outside fades, replaced by swirling memories and fragmented flashes of light that echo her innermost fears and hopes. The sound of distant water dripping morphs into a heartbeat, pulsating with each revelation as the camera dives into the abyss of her thoughts, revealing a dreamscape rife with echoes of past choices and future uncertainties.
Thank you for reading.
Have a creative week!
