Today:

• OpenAI’s New Sora App Blends AI Video with Human Presence

• Google’s New AI Mode Lets You Search with Photos and Prompts

• Microsoft’s Copilot Now Talks with a Face and a Voice

• Former OpenAI, DeepMind Talent Launches Periodic Labs to Reinvent Science

• Meta’s Code World Model Thinks Like a Real Developer

OpenAI’s New Sora App Blends AI Video with Human Presence

Sora 2, OpenAI’s new video and audio generator, creates realistic scenes with accurate physics, dialogue, and sound. It powers a new iOS social app that lets users insert themselves into AI-generated videos. The app prioritizes creativity, safety, and user control.

KEY POINTS

• Realism and Control: Sora 2 improves physics, motion, and world consistency in generated video, including failure modeling—like a basketball missing the hoop and rebounding.

• Social App with Cameos: The new Sora app lets users add themselves into AI-generated scenes via a one-time likeness capture—ushering in a new way to communicate with friends.

• Safety and Ethics: OpenAI built tools for feed control, consent, teen protection, and moderation, aiming to avoid addictive or harmful use patterns.

Why it matters

Sora 2 shows how far AI has come in understanding and simulating the real world. This matters because the ability to model reality accurately is key to future AI agents and robotics. But more than that, it opens a new way to create and share stories, safely and socially—where anyone can be the star of a scene. OpenAI is trying to make this powerful tech fun, creative, and safe from the start.

Google’s New AI Mode Lets You Search with Photos and Prompts

Google just launched a smarter AI Mode in Search that helps you explore visually by chatting naturally, uploading images, and getting rich results you can shop or learn from. It understands your intent better than ever, even from vague questions or photos.

KEY POINTS

• Visual + Conversational Search: You can now use images, casual descriptions, or follow-up questions to explore design ideas or products with ease.

• Smarter Shopping: Skip filters—just describe what you want (like “jeans that aren’t too baggy”), and AI Mode shows curated, shoppable results from over 50 billion listings.

• Context-Aware AI: Google’s “visual search fan-out” uses advanced image analysis to understand what’s in a photo and deliver results with more depth and relevance.

Why it matters

This update makes Google Search more helpful when words fail. Whether you're decorating a room, planning a wardrobe update, or just browsing ideas, AI Mode makes the process feel more like talking to a helpful friend than using a search engine. It's a big step toward more natural, intuitive technology.

Microsoft’s Copilot Now Talks with a Face and a Voice

Microsoft’s Copilot now offers 40 animated human avatars called “Portraits” that show expressions and lip-sync during voice chats, aiming to make AI feel more friendly and natural. The feature is available to limited users in the US, UK, and Canada.

KEY POINTS

• Real-Time Animated Faces: The new Portraits feature adds expressive human avatars to voice-based Copilot interactions using Microsoft's VASA-1 tech.

• Emotional Comfort: Microsoft says users feel more at ease talking to a face, making AI conversations feel more like human ones.

• Safety Measures: Avatars are stylized (not photorealistic), with usage limits, clear AI labels, and 18+ age restrictions for safety and transparency.

Why it matters

Talking to AI is becoming more like talking to people. By adding realistic faces and expressions, Microsoft is making Copilot feel friendlier and easier to use—especially for voice conversations. It’s part of a broader trend toward humanizing AI while still being careful about safety and ethics.

🧠RESEARCH

SLA: Beyond Sparsity in Diffusion Transformers via Fine-Tunable Sparse-Linear Attention

SLA is a new attention method that speeds up video-generating AI models by 13× without losing quality. It smartly focuses computing power on only the most important data. By combining two types of efficient shortcuts and fine-tuning existing models, it cuts attention costs by 95% and boosts overall performance.

StableToken: A Noise-Robust Semantic Speech Tokenizer for Resilient SpeechLLMs

StableToken is a new speech tokenizer that keeps token sequences consistent even in noisy audio. Unlike older models that break under minor sound changes, it uses multiple audio paths and a voting system to stay stable. This helps speech-based AI models understand better and perform more reliably in real-world conditions.

OpenGPT-4o-Image: A Comprehensive Dataset for Advanced Image Generation and Editing

OpenGPT-4o-Image is a new dataset built to train AI models in complex image generation and editing. It includes 80,000 instruction-image pairs across 11 domains like science diagrams and multi-step edits. By organizing tasks and automating data creation, it boosts model performance by up to 18% on editing and 13% on generation benchmarks.

🛠️TOP TOOLS

Each listing includes a hands-on tutorial so you can get started right away, whether you’re a beginner or a pro.

Explain Paper – AI Research Assistant -browser‑based AI reading companion for scholarly PDFs

Songtell – AI Lyrics Interpreter / Music Analysis Tool - AI‑powered platform that “unravels the stories behind the lyrics you love,” providing structured explanations of songs

DeepSwap AI – AI Face‑Swap & Video Editor -enerative‑AI app for swapping faces in videos, photos, and GIFs

📲SOCIAL MEDIA

🗞️MORE NEWS

• Ex-OpenAI and DeepMind researchers raised $300M to launch Periodic Labs, aiming to build robot-run labs that discover new materials like superconductors and generate fresh data to train smarter AI scientists.

• Meta’s new Code World Model learns how code behaves, not just how it looks. Trained on execution traces and simulated coding tasks, it outperforms peers on key benchmarks by reasoning like a real developer.

• CoreWeave signed a $14.2 billion deal with Meta to supply AI computing power through 2031. This boosts CoreWeave’s growth beyond Microsoft, strengthens Meta’s AI push, and fuels debate over an AI infrastructure bubble.

• Character.AI removed Disney characters like Elsa and Darth Vader after Disney sent a legal warning over copyright violations. The move highlights growing tension over AI-generated content, intellectual property, and child safety on chatbot platforms.

• Amazon launched four new Echo devices powered by Alexa+, featuring smarter conversations, custom AI chips, and advanced sensors. The devices support health tracking, smart home automation, and immersive entertainment through upgraded sound and displays.