Wispr Flow - Don’t type, just speak
Mini toolsOVIE
By Kyutai
OVIE, short for “One View Is Enough!”, is Kyutai’s framework for monocular novel view generation from single images. The repository describes it as a model trained entirely on unpaired internet images rather than multi-view image pairs, making it a different supervision setup from many classic view-synthesis systems. It is positioned for in-the-wild novel view synthesis, with official code, pretrained weights on Hugging Face, and local checkpoints for evaluation and training. The repo’s inference examples show a pretrained OVIEModel generating new views from one RGB image plus camera pose inputs.
New
Image
Gen 3
Overview
OVIE is Kyutai’s open monocular novel view synthesis model for generating new viewpoints from a single in-the-wild image. It is trained without paired multi-view supervision, relying instead on unpaired internet images, and is aimed at 3D-aware view generation from just one input view.
About Kyutai
Kyutai is a non-profit laboratory dedicated to open research in artificial intelligence. Our team is focused on few, high-impact projects at the frontier of current knowledge in generative AI. Join us to invent the next generation of AI algorithms and share them with the world!
Company Size:
20
Website:
kyutai.org
Last updated: April 15, 2026
