# AI NEWS: 28 Releases and Updates You Missed This Week Source: https://www.youtube.com/watch?v=vpUk5OetUNk Recap page: https://rapidrecap.app/video/vpUk5OetUNk Generated: 2025-08-23T13:35:25.162+00:00 --- ## Quick Overview Nvidia released the Seed-OSS-36B AI model with 512K token context, featuring toggleable on/off reasoning, and Google announced Gemini for Home, bringing AI to Google Home devices with enhanced features like voice translation and AI-powered coaching. **Key Points:** - Nvidia released Seed-OSS-36B, an open-source AI model with 685B parameters and a 512K token context window, featuring toggleable on/off reasoning. - Google announced Gemini for Home, bringing AI to Google Home devices with voice translation and personalized coaching. - Google's AI Mode in Search now includes agentic capabilities for tasks like restaurant reservations and event ticket booking. - Runway launched its Game Worlds Beta, allowing users to create text-based games with AI-generated stories and characters. - ElevenLabs introduced a Video-to-Music flow in its Studio, enabling custom soundtrack generation based on video context. - Grammarly is rolling out new AI agents for writing assistance, including plagiarism detection and grading, at no extra cost for Free and Pro users. - Meta is rumored to release smart glasses with an integrated display and AI features in September, starting at $800. ![Screenshot at 00:00: The presenter introduces the week's AI news roundup, highlighting several significant updates from major tech companies and AI projects.](https://ss.rapidrecap.app/screens/vpUk5OetUNk/00-00-00.png) **Context:** This week in AI news, several major companies and projects announced significant updates and new features. Nvidia released a new large language model, Google is integrating Gemini across its home devices and search, Runway launched new AI game world features, ElevenLabs introduced a video-to-music tool, Grammarly is rolling out AI agents for writing assistance, and Meta is reportedly preparing to release smart glasses with AI features. ## Detailed Analysis This week saw significant AI developments, starting with Nvidia's release of a new open-source Seed-OSS-36B model, boasting 685 billion parameters and a 512K token context window, allowing for on/off reasoning toggles and improved performance across various benchmarks. Following this, Google announced Gemini for Home, an AI assistant designed to integrate with Google Home devices, offering features like real-time voice translation and personalized AI coaching. Google's AI Mode in Search also received updates, introducing new agentic capabilities for tasks like restaurant reservations and event ticket booking. Furthermore, ElevenLabs launched a Video-to-Music flow in its Studio, enabling users to generate custom soundtracks based on video context and layer voiceovers and sound effects. Grammarly is introducing new AI agents for specific writing challenges, such as plagiarism detection and grading, available for Free and Pro users at no extra cost. Finally, Meta is reportedly set to unveil smart glasses with a display in September, featuring on-lens alerts and wrist accessory controls, with the AI potentially assisting in tasks like finding the right tool. ### Nvidia Releases Seed-OSS-36B Model - 685B parameters, 512K context window, toggleable reasoning - Available on Hugging Face ### Google Announces Gemini for Home - AI assistant for smart home devices, features voice translation, AI coaching, integrations with Google apps ### Google AI Mode Updates - New agentic capabilities for tasks like restaurant reservations and event tickets ### ElevenLabs Launches Video-to-Music Flow - Generates custom soundtracks from video context, allows voiceover and SFX layering ### Grammarly AI Agents - New features for plagiarism detection, grading, and AI-native writing surface ### Meta Smart Glasses - Display in September, on-lens alerts, wrist accessory controls, AI for identifying tools ### GPT-5 Development - OpenAI CEO Sam Altman emphasizes user-driven features like memory and personality ![Screenshot at 27:42: Nvidia's Seed-OSS-36B model with 512K token context, highlighting its on/off reasoning capability.](https://ss.rapidrecap.app/screens/vpUk5OetUNk/00-27-42.png) ![Screenshot at 27:02: DeepSeek V3.1 benchmarks showing improved performance across various tasks compared to previous versions.](https://ss.rapidrecap.app/screens/vpUk5OetUNk/00-27-02.png) ![Screenshot at 15:07: Runway Game Worlds Beta announcement, showcasing AI-generated text-based games.](https://ss.rapidrecap.app/screens/vpUk5OetUNk/00-15-07.png) ![Screenshot at 11:16: Runway AI's Gen-4 Image Turbo enabling users to generate images with reference images.](https://ss.rapidrecap.app/screens/vpUk5OetUNk/00-11-16.png) ![Screenshot at 08:11: Demonstration of Gemini's real-time voice translation feature during a phone call.](https://ss.rapidrecap.app/screens/vpUk5OetUNk/00-08-11.png) ![Screenshot at 05:01: Nvidia's AI model for coloring and upscaling black and white photos.](https://ss.rapidrecap.app/screens/vpUk5OetUNk/00-05-01.png) ![Screenshot at 02:21: Examples of image editing using AI, showing the transformation of a person into cartoon styles.](https://ss.rapidrecap.app/screens/vpUk5OetUNk/00-02-21.png) ![Screenshot at 00:00: The presenter introduces the AI news roundup for the week.](https://ss.rapidrecap.app/screens/vpUk5OetUNk/00-00-00.png)