OmniShowcase: Omnidirectional Subject-Consistent
Video Generation from Multi-View Images

Junyeong Ahn^1*, Geonwoo Kim^2*, Kinam Kim^1*, Kihwi Kim³, Hoonjin Jung³, Hyojin Jang¹, Jaegul Choo¹

¹KAIST AI ²Inha University ³FLIPTION

Paper Code Dataset

Orbit 360°

The video begins with a snowman wearing a wizard hat and a black cloak, hugging a small bird, standing on a floating ice block in a serene, icy landscape with a backdrop of glaciers and snow-capped mountains. The video concludes with a frontal view of the snowman and the bird, highlighting the peaceful and picturesque setting. + [Orbit 360°]

Abstract

Generating videos of a specific subject from a few reference images remains challenging when the camera reveals surfaces that are not visible in the input. Early subject-to-video (S2V) methods condition on a single reference image and must therefore infer the subject's unseen appearance and geometry as the viewpoint changes. Recent methods extend S2V to multiple reference views, providing more information about the subject from different viewpoints, but they do not explicitly specify the 3D structure that connects these observations. In addition, existing training data do not guarantee that viewpoint changes preserve the same underlying 3D subject. We present OmniShowcase, built on two components. First, a fully automated data synthesis pipeline produces training videos along diverse camera trajectories, with viewpoint variation driven by underlying subject geometry, even spanning full 360-degree orbits. Second, a geometry-conditioning method provides the video model with a turntable of surface-normal maps reconstructed from the reference views, without requiring temporal alignment with the generated frames. We construct Showcase-4K, a corpus of 3D-asset-based videos covering twelve camera motions across 88 trajectory configurations with reconstruction-derived surface-normal conditioning, and ShowcaseBench, a held-out evaluation set that ships the source meshes and camera parameters used to render each video. Experiments show state-of-the-art subject identity and 3D consistency while maintaining competitive video quality and motion.

Method

Four reference views are reconstructed in 3D and rendered as a surface-normal turntable. References and normals are concatenated with the target span and read in context by a LoRA-tuned image-to-video diffusion transformer.

Automated 3D-consistent data synthesis pipeline — **Data synthesis pipeline.** Blender renders the target orbit and four reference views from one curated 3D asset. The target is composited into a prompt-driven scene, and the four views are reconstructed and rendered as the normal turntable used as the condition.

Architecture overview — **Architecture.** The reference images and the normal condition enter the reference pathway of the diffusion transformer, with no frame-wise alignment between the normal turntable and the generated video.

Beyond the turntable

The normal condition is a fixed yaw sweep, but the generated camera is not bound to it. The target follows whichever trajectory the prompt names.

Arc-Up R→C

The video begins with the wooden teddy bear statue standing on a hand-painted blue-and-yellow majolica tile ledge on an Amalfi Coast terrace at bright midday, surrounded by lemon trees heavy with yellow lemons, with pastel yellow and coral houses stacked on the cliffside and a sparkling turquoise sea below, every leaf and tile in crisp sharp focus. + [Arc-Up R→C]

Comparison with baselines

Same references, same prompt and same trajectory for every method.

ShowcaseBench

Arc-Down L→C

Phantom

MAGREF

MV-S2V

3DreamBooth

Ours

The video begins with the wooden teddy bear statue standing small on the wide wooden porch of a snow-covered log cabin at dusk, seen from a distance so the whole porch is visible, with glowing string lights along the railing, a stack of firewood, a lantern by the door, tall snow-laden pine trees, and warm light spilling from a frosted window behind. + [Arc-Down L→C]

NAVI

Orbit 360°

Phantom

MAGREF

MV-S2V

3DreamBooth

Ours

The video starts with a small decorative ice cream cart model with a green base and a red and white striped canopy, set on the painted ledge of a fairground arcade stall. + [Orbit 360°]

Qualitative results

Qualitative comparison with baselines — Each method carries three marks for whether the camera follows the trajectory the prompt names (Traj.), whether the subject keeps its identity (ID), and whether it stays 3D-consistent (3D).

Quantitative results

Bold is best, underline is second best.

	Method	Identity		3D consistency			Text	Video quality
	Method	DINO↑	CLIP↑	MEt3R↓	GT-CD↓	RotCov↓	ViCLIP↑	Imaging↑	Aesthetic↑	Dynamic↑
Held-out	Phantom-MV	0.621	0.871	0.444	0.2843	0.906	0.236	72.52	0.595	0.460
	MAGREF-MV	0.616	0.876	0.409	0.1789	1.177	0.242	73.88	0.614	0.760
	MV-S2V	0.688	0.896	0.351	0.2306	0.787	0.244	73.21	0.590	0.960
	3DreamBooth	0.707	0.907	0.325	0.1717	0.765	0.238	70.67	0.598	1.000
	OmniShowcase	0.799	0.925	0.190	0.0824	0.665	0.259	74.29	0.574	1.000
NAVI	Phantom-MV	0.690	0.882	0.462	0.2014	1.458	0.255	71.17	0.572	0.029
	MAGREF-MV	0.665	0.892	0.457	0.2145	1.284	0.251	69.81	0.574	0.343
	MV-S2V	0.770	0.912	0.350	0.1816	1.163	0.250	71.73	0.565	1.000
	3DreamBooth	0.741	0.912	0.386	0.2272	1.179	0.244	70.34	0.559	1.000
	OmniShowcase	0.828	0.918	0.249	0.0975	1.074	0.260	74.58	0.551	1.000

Ablation

	Geometry condition	Identity		3D consistency			Text	Video quality
	Geometry condition	DINO↑	CLIP↑	MEt3R↓	GT-CD↓	RotCov↓	ViCLIP↑	Imaging↑	Aesthetic↑
Held-out	Normals (ours)	0.799	0.925	0.190	0.0824	0.665	0.259	74.29	0.574
	None	0.752	0.903	0.265	0.2092	0.727	0.248	69.85	0.547
	Textured	0.755	0.910	0.270	0.1698	0.747	0.236	73.41	0.564
	Sparse	0.752	0.912	0.263	0.1777	0.698	0.250	71.84	0.552
NAVI	Normals (ours)	0.828	0.918	0.249	0.0975	1.074	0.260	74.58	0.551
	None	0.793	0.904	0.315	0.1987	1.107	0.238	67.67	0.532
	Textured	0.780	0.907	0.323	0.1842	1.089	0.240	71.47	0.544
	Sparse	0.772	0.909	0.288	0.1769	1.090	0.240	67.72	0.531

OmniShowcase: Omnidirectional Subject-ConsistentVideo Generation from Multi-View Images