Clones camera motion from a reference video onto a new source image, including multi-shot cuts, dolly zooms, bullet time and fisheye effects. No code or models released yet.
Research model that clones multi-shot camera motion from a reference video onto a new source image.
Updated EU status checked Last shown
OmniDirector at a glance
- What is OmniDirector?
- OmniDirector is a model from Kling. Research model that clones multi-shot camera motion from a reference video onto a new source image.
- Who makes OmniDirector?
- OmniDirector is made by Kling.
- Does OmniDirector work in the EU?
- Unverified. No explicit evidence found; the model is a research project with no public release or access. Last verified 14 Nov 2026.
- How much does OmniDirector cost?
- Pricing unknown. No public access, API or pricing found; code and models not released.
- Where was OmniDirector shown?
- In 1 creator video by AI Search. Most recently by AI Search on 21 Jun 2026, at 3:54 in “New robot waifus, GLM 5.2 craze, AI spas, new world models, new science agents: AI NEWS”.
Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
OmniDirector is a research project, presented in the paper "OmniDirector: General Multi-Shot Camera Cloning without Cross-Paired Data". It transfers camera motion from a reference video to a new source image to generate video. The authors are associated with Kling.
According to a creator review, it can reproduce multi-shot cuts and effects such as dolly zooms, bullet time and fisheye looks. The method's title indicates it is trained without cross-paired data, meaning it does not need paired examples of the same content filmed with different camera paths.
The project page currently shows only generated example clips and a citation. As of the creator coverage, no code or model weights had been released, and no public product, API or pricing was found.
- Clones camera motion from a reference video onto a new source image
- Supports multi-shot sequences with cuts
- Reproduces effects such as dolly zoom, bullet time and fisheye
- Trained without cross-paired data
As seen on YouTube
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
Similar AI tools
Closest neighbours by meaning, not just by category.
MoCam
Model
Research method for re-rendering an existing video along a new camera trajectory while preserving subject motion.
Kling
Model · Kuaishou
Kuaishou's AI video and image generation model family, offered via the Kling AI web studio and API.
Kling 4
Model · Kuaishou
Kuaishou's Kling 4 video generation model with multi-reference input, keyframe control, native stereo audio and HDR output.
FastH3 Omni
Model · Hao AI Lab
Distilled MiniMax H3 reference-to-video-and-audio model from Hao AI Lab, generating in 8 steps instead of 50.
Krea 2
Model · Krea
Krea's text-to-image model, offered on the Krea creative platform; compared by creators with open local image models.
Flex4DHuman
Research
Multi-view video diffusion model that turns monocular or sparse videos of dynamic humans into dense multi-view videos for 4D Gaussian splats.