Kinetix Unveils Kamo-1: A 3D-Conditioned AI Video Generation Model with Physical Grounding
Kinetix introduces Kamo-1, a 3D-conditioned video generation model that brings physical grounding and precise control to AI video generation. The model supports multimodal conditioning including 3D camera trajectories, 3D character animations reconstructed from RGB acting videos, reference images, and natural language prompts. It enables coordinated control over both camera motion and character motion within the same generated shot, representing a research advancement in 3D and human motion intelligence.
Key quotes
Kamo-1 is our 3D conditioned video generation model, designed to bring physical grounding and precise control into AI video.
It supports multimodal conditioning, including 3D camera trajectories, 3D character animations reconstructed from RGB acting videos, reference images, and natural language prompts.
By jointly conditioning on both camera trajectories and 3D animations, Kamo-1 enables coordinated control over camera motion and character motion within the same generated shot.
From the article
You might also wanna read
LingBot-Video: A Video Generator for Robotics
ByteDance's Seedance 2.0 Shows Progress in AI Video Generation but Still Has Limitations
The article discusses ByteDance's Seedance 2.0, a new AI video generation model that shows impressive improvements in creating realistic vid
The Verge·4mo ago
Kimi K2.5: Open-Source Multimodal AI Model with Visual Agentic Intelligence and Agent Swarm Capabilities
Kimi K2.5 is introduced as the most powerful open-source model to date, building on Kimi K2 with continued pretraining on approximately 15 t
Lumos-Nexus: A Training-Efficient Two-Stage Framework for High-Fidelity Video Generation with Reasoning Capabilities
Lumos-Nexus is a training-efficient unified video generation framework that addresses the computational challenge of integrating large high-
arxiv.org·1mo ago
Helios: A 14B Parameter Real-Time Video Generation Model for Minute-Scale Content
Helios is a 14B parameter video generation model that achieves real-time performance at 19.5 FPS on a single NVIDIA H100 GPU while supportin
alphaxiv.org·4mo ago
BilliardPhys-Bench: New Benchmark Reveals Physical Reasoning Gaps in Multimodal AI Models
This paper introduces BilliardPhys-Bench, a benchmark designed to evaluate multimodal large language models (MLLMs) on intuitive physical re
arxiv.org·1mo ago
Comments
Sign in to join the conversation.
No comments yet. Be the first.