Quanta AI Labs

11 / Mar 10 – 16, 2025 · 10 RELEASES

AI News Week 11, 2025

Week 11 2025 AI releases.

Video & Animation (3)

  • DreamActor-M1

    ByteDance Intelligent Creation · Video Editing

    A diffusion transformer-based framework for human image animation that uses hybrid guidance (facial representations, 3D skeletal data, and visual references) to achieve fine-grained control over expressions and movements across multiple scales with long-term temporal coherence.

  • TTT Video DiT

    test-time-training · Text-to-Video

    Test-time training framework for video diffusion transformers that adapts models at inference time to improve generation quality and consistency. Enables dynamic optimization of video generation without retraining.

  • Video DiT

    Test-Time Training · Text-to-Video

    A diffusion transformer model for video generation that applies test-time training to improve generation quality and coherence. Leverages adaptive computation during inference to enhance video synthesis performance.

3D & Spatial (2)

  • GenXD

    National University of Singapore & Microsoft · 3D Generation

    GenXD is a unified framework for generating 3D and 4D scenes that disentangles camera and object movements using multiview-temporal modules. The work introduces CamVid-30K, a large-scale real-world 4D dataset with 30K videos annotated with camera poses and object motion.

  • Hi3DGen

    ByteDance, CUHK, Tsinghua University · 3D Generation

    A framework for generating high-fidelity 3D geometry from 2D images using normal map estimation as an intermediate representation, employing noise-injected regression and normal-regularized latent diffusion to capture fine geometric details.

Developer Tools (4)

  • WhatsApp MCP

    lharries · AI Infrastructure

    An MCP (Model Context Protocol) server that enables Claude and other LLMs to read and send WhatsApp messages programmatically. Provides integration between AI assistants and WhatsApp messaging for automation and communication workflows.

  • Self.so

    Self.so (Powered by Together.ai) · Prompt Engineering

    An open-source tool that converts resumes or LinkedIn profiles into professional websites in under a minute using AI. Built on Together.ai's Qwen model for automated resume parsing and webpage generation.

  • MCP Servers

    Glama · AI Infrastructure

    Model Context Protocol (MCP) servers enable standardized connections between AI applications and data sources through a unified interface for tool use and resource access.

  • GitMCP

    GitMCP · AI Infrastructure

    GitMCP creates instant Model Context Protocol servers for any public GitHub repository, allowing AI assistants to understand code context by automatically parsing readme.md, llms.txt, and other repo files. Works with Claude, Cursor, Windsurf, and other MCP-compatible AI tools through simple domain replacement.

Research & Safety (1)

  • Tracing Thoughts

    Anthropic · Interpretability

    Research on interpretability techniques for understanding and visualizing the internal reasoning processes of language models through thought tracing mechanisms.

Published by Quanta AI Labs, an AI agency in India. Browse every week in AI News Weekly.