TAAFT
Free mode
100% free
Freemium
Free Trial
Deals

SANA WM (Bidirectional)

By NVIDIA
SANA-WM Bidirectional is an Apache 2.0 image-to-video diffusion world model from Efficient-Large-Model. The released checkpoint is a 2.6B-parameter diffusion transformer trained for efficient one-minute generation, producing 720p videos from an input image and prompt while supporting precise six-degree-of-freedom camera control. Its design uses hybrid linear attention for long-context modeling, dual-branch camera control for trajectory adherence, a two-stage generation and refiner pipeline for temporal consistency, and a camera-pose annotation pipeline for spatiotemporally consistent supervision.
New Multimodal Gen 3
Released: May 14, 2026

Overview

SANA-WM Bidirectional is Efficient-Large-Model’s open-source 2.6B image-to-video world model for 720p minute-scale video generation with 6-DoF camera control.

About NVIDIA

Industry: Computer Hardware Manufacturing
Company Size: 42000
Location: Santa Clara, California, US
Website: nvidia.com
View Company Profile

Tools using SANA WM (Bidirectional)

No tools found for this model yet.

Last updated: May 22, 2026
0 AIs selected
Clear selection
#
Name
Task