VideoAtlas and 3D Work in Blender
The open Atlas boasts a cosmic view. Non-photorealistic renders and Blender restoration inspire awe.
Это не новости, это персональный экспертный взгляд на то, как развивается индустрия ИИ, графики, метаверса, крипты, нейротехнологий и в каком направлении катится все это безобразие. Для связи: @SergTsyp
Open in Telegram
VideoThe open Atlas boasts a cosmic view. Non-photorealistic renders and Blender restoration inspire awe.
VideoWorld Labs released the Atlas model with an open API supporting 3D reconstruction, video rendering, and export to 3D Gaussian Splats for games and simulators.
VideoAnthropic launched Claude Meetings for AI-assisted video calls: collaborative editing, data access, and automatic notes in beta.
An automated portal for the NeuroRoast archive has been developed: AI agents collected data, uploaded 347 videos to YouTube, and created a lightweight site with analytics.

AntiGravity released an unscheduled update: hydrogen is now available to Pro subscribers. Details on limits will be published tomorrow.
Video
VideoI seem to have messed things up. I'll show you how. As for the roast today... umm.. roasting. Sunsets in Paphos are the best. Water 24, air 24.

Analysis of CEO Huang's dismissal: how personal life affects business, the role of security services, and Elon Musk's example of independence.
VideoNVIDIA Research released Déjà View (117M parameters), a compact model for photo-based 3D reconstruction. The architecture uses recurrent transformers for a flexible balance between speed and accuracy.

Zyphra released the open Zonos3-Mini model for voice cloning in 43 languages. The model runs in the browser via WebGPU without servers.
VideoLuma AI has released the Character Swap function for replacing and adding objects in videos via Dream Machine and the Luma MCP plugin.

Google released Gemini Decision 1 for content auditing at a price of $0.10 per 1M tokens. The model supports native analysis of video, audio, and PDFs with a 256K token context.
VideoBytePlus launched the AI model Dreamina Seedance 2.5 Flash for rapid video generation. Speed increased twofold, and price reduced by 34%. API access opens on November 15.
VideoHow to create a gangster series using AI: a breakdown of the process for generating the final shootout scene in Seedance 2.5 and project costs.
VideoI'm a bit tired, wrote some news, tested Minimax with images, and made a collaborative notebook for her. Now I'll fry up something delicious and have a beer with the girls. I'm not Huang, so it's allowed.
Video
VideoSpaghetti Splatting is a generalization of 3DGS to curvilinear structures. The method ensures line continuity, built-in anti-aliasing, and data readiness for physical simulation of dynamics.
VideoStripe has completed the purchase of OpenRouter, an AI model aggregator. The platform unites over 400 models and processes 10 trillion tokens daily to optimize AI expenses.
Overview of winners in the FLUX and MiniMax partner nominations. Open-source tools for video generation, 3D editing, and audio post-production.
VideoResults of the #OpenSourceHiggsfield hackathon: 300 developers from 45 countries replicated features of the closed Higgsfield platform. Overview of top projects and winners.

Alibaba released Qwen4-Max-Preview with a 1M token context and built-in agent mode for complex coding and autonomous task planning.
VideoNuance Labs released the Nuplex model and the Emma avatar for real-time dialogues without the delays characteristic of traditional speech synthesis systems.

The Fable 5.2 model created a photorealistic scene from scratch by writing code for a tracer and geometry generation. The result is available on GitHub.

Apple is preparing the Harbor app for macOS, which allows combining the memory of several Macs via Thunderbolt 5 to run heavy AI models locally.