CopyQuad: Quadrupedal Skill Transfer from a Single Side-View Video Without a Depth Model

Mohamad H. Danesh; Raina Hu; Mengxue Xu; Hsiu-Chin Lin

cs.ROOct 12, 2026

CopyQuad: Quadrupedal Skill Transfer from a Single Side-View Video Without a Depth Model

Authors: Mohamad H. Danesh, Raina Hu, Mengxue Xu, Hsiu-Chin Lin

Organizations: McGill University · Mila – Quebec AI Institute

Abstract

Video-to-robot pipelines for quadrupedal imitation insert a learned lifting stage, either a monocular depth model or a per-clip 3D reconstruction, to recover the one axis a side-view camera cannot see. We present CopyQuad, which omits that stage and pins the missing axis to the robot's own stance width instead. The setting is side-view footage of sagittally dominated motion, where limb motion is almost entirely in the image plane. A frozen 2D keypoint detector and a fixed analytic retargeting then turn a single uncalibrated clip into a robot reference whose foot targets are reachable by construction, so no per-clip trajectory optimization is needed. One phase-indexed reward structure, whose only per-behavior freedom is three scalars from a shared sweep, learns a dog walk, a horse walk, a horse trot and two non-periodic gestures, all deployed zero-shot on hardware. We then measure what omitting the axis costs. In a training-free rigidity test, a video depth model varies the distance between the animal's hips by $40$ to $107\%$ of its own mean and a single-image metric model by $22$ to $58\%$ , though that distance cannot move. Motion capture puts the true lateral paw travel of a dog walk at $4.5$ cm peak to peak, a mean deviation near $1.4$ cm against the $3.5$ cm of lateral foot error our policies already leave, so the discarded axis sits below the pipeline's resolution. Depth repaired to feasibility never beats the constant, and the unrepaired estimates put the robot on the ground.

Figures & tables

Fig. 1 : CopyQuad. A single monocular clip becomes a robot-ready reference through one frozen 2D keypoint detector followed by a chain of fixed analytic steps, namely sagittal-plane lifting with the lateral axis pinned to the robot’s stance width, kinematic retargeting, and root recovery. A phase-indexed policy $\bm{\pi}_{\psi}$ trained in simulation tracks that reference and is deployed zero-shot on hardware. Only the reference itself and three reward scalars change between behaviors.

Fig. 2 : The reference quantities. Left : the stance tracks $y_{k}^{\text{nom}}{=}{\pm}w/2$ ( dashed ) that Eq. 1 pins the lateral axis to, with the base $\hat{\mathbf{x}}$ and speeds $v_{x}$ , $v_{y}$ . Middle : the robot height $H_{r}$ , a foot $\mathbf{p}_{\ell}$ and the stance sweep $\bar{d}$ . Right : the detector keypoints , all that is read from pixels, and the spans $L_{a}$ , $H_{a}$ that Eq. 2 matches to $L_{r}$ , $H_{r}$ .

Fig. 3 : Source video $\to$ Isaac Lab $\to$ real robot for the two gesture clips. All four feet are tracked, but only one front and one hind are ringed, $\bullet$ front and $\bullet$ hind, since left and right overlap in a side view.

Fig. 4 : Source video $\to$ Isaac Lab $\to$ real robot for the two gesture clips. The gestures are read from the near-side pair alone, so only two feet are ringed, $\bullet$ front and $\bullet$ hind.

Source clip		Dog walk	Horse walk	Horse trot	Dog bow	Dog give-paw
	fps, length (s)	30, 21.4	25, 21	30, 36	50, 7	30, 8
Simulation	$\hat{\bar{v}}\to\bar{v}$ (m/s)	$0.29\to 0.29$	$0.48\to 0.55$	$0.58\to 0.58$	$0.00\to 0.00$	$0.00\to 0.00$
	$e_{q}$ per-clip (deg) $\downarrow$	$14.3{\scriptstyle\pm 1.4}$	$13.3{\scriptstyle\pm 1.7}$	$11.6{\scriptstyle\pm 1.3}$	$4.7{\scriptstyle\pm 0.3}$	$8.3{\scriptstyle\pm 4.4}$
	$e_{q}$ shared (deg) $\downarrow$	$18.1{\scriptstyle\pm 1.6}$	$15.7{\scriptstyle\pm 2.1}$	$13.4{\scriptstyle\pm 1.9}$	$4.2{\scriptstyle\pm 0.4}$	$4.6{\scriptstyle\pm 0.7}$
	$e_{p}$ (cm) $\downarrow$	$13.7{\scriptstyle\pm 1.8}$	$15.0{\scriptstyle\pm 2.6}$	$15.5{\scriptstyle\pm 2.0}$	$7.4{\scriptstyle\pm 0.4}$	$11.8{\scriptstyle\pm 6.8}$
	$e_{p}$ fore-aft / lateral / vert. (cm)	$9.8/3.5/3.7$	$10.5/3.8/4.0$	$11.0/3.9/4.2$	—	—

TABLE I : All five behaviors on ANYmal-D, from the source clip through simulation to zero-shot hardware.

e_{q}

per-clip and

e_{q}

shared differ only in the reward weights, namely three per-clip scalars or one fixed vector for all five, and both are evaluated against the same full-method reference, so the per-clip row is the full bar of Fig. 5 . A dash marks a quantity undefined for a stationary gesture.

	$e_{q}^{\mathrm{ref}}$ (deg) $\downarrow$	$\bar{v}$ (m/s)	Surv. (s)
CopyQuad (ours)	$14.3{\scriptstyle\pm 1.4}$	$0.29$	$19.9{\scriptstyle\pm 0.2}$
Replacing the lifting stage
SLoMo [ 8 ] †	$33.6-36.6$	${\approx}0$	—
SLoMo+physics refinement [ 8 , 26 ] †	$19.5{\scriptstyle\pm 1.1}$	$0.22{\scriptstyle\pm 0.1}$	$20.0$
Replacing the objective
AMP [ 27 ] †	—	$0.03$	—

TABLE II : Alternatives on the dog walk. Each row replaces one stage and leaves the rest untouched, the upper group the lifting stage and the lower group the objective. All use

e_{q}^{\mathrm{ref}}

against the full method’s reference, survival capped at

20

s. Dashes are undefined quantities. † marks the need to an extra conversion (SLoMo) or a data regime we do not supply (AMP).

	VDA (relative)		Depth Pro (metric)
Clip	$\|\Delta z_{\mathrm{c}}\|$	CV	$\|\Delta z_{\mathrm{c}}\|$ (cm) / CV	sign
Dog walk	$2.42$	$40\%$	9 / $22\%$	98%
Horse walk	$0.00$	—	6 / $35\%$	86%
Horse trot	$0.01$	—	4 / $58\%$	61%
Dog bow	$1.68$	$75\%$	11 / $28\%$	95%
Give-paw	$0.94$	$107\%$	10 / $31\%$	94%

TABLE III : The rigidity test. Implied left-right hip separation

|\Delta z_{\mathrm{c}}|

along the camera’s depth axis, subscripted to distinguish it from the robot’s vertical

z

, under a relative video model (VDA [ 28 ] , model units) and a single-image metric model (Depth Pro [ 29 ] , cm). CV is blank on the horse clips, where VDA leaves no separation to take a ratio of.

Fig. 5 : Retargeting ablation on the dog walk and the horse trot. Bars are

e_{q}^{\mathrm{ref}}

against the full method’s reference. Every variant is evaluated on one fixed target, with whiskers

\pm 1

std over seeds. Color marks what the policy did, not how well it tracked. Blue survives the horizon and locomotes. Red falls before

15

s, so its

e_{q}

is post-fall accumulation rather than tracking, with survival annotated. Gray stays up but never locomotes, so a low

e_{q}

is a stationary robot.

	MoCap route	Video route (ours)
$e_{q}$ (deg) $\downarrow$	$9.5{\scriptstyle\pm 1.0}$	$12.5{\scriptstyle\pm 1.4}$
$e_{p}$ (cm) $\downarrow$	$10.5{\scriptstyle\pm 1.5}$	$13.5{\scriptstyle\pm 1.8}$
Surv. (s, cap 20)	20.0	20.0

TABLE IV : What video costs against MoCap, on one dog walk taken from markers and from a side-view rendering of the same motion. Both routes share the retargeter, reward and budget, and both are evaluated against the marker-derived reference.

Explore similar work

Unleashing Infinite Motion: Scaling Expressive Quadrupedal Motion via Generative Video Priors
Jun 26, 2026Youzhi Liu, Li Gao, Yifei Qian +3Synthetic Data Generation for Robotics Robot Motion Generation
Amap, Alibaba Group
Learning Quadruped Walking from Seconds of Demonstration
Mar 7, 2026Ruipeng Zhang, Hongzhan Yu, Ya-Chien Chang +3Robot Imitation Learning Data-Efficient Robot Learning
University of California, San Diego
MimicAgent: Quadruped Skills via Prompt-to-Trajectory Generation
Sep 21, 2026Lucky Kant Nayak, Narayanan Palghat Parameswaran, Neehar Peri +1Reinforcement Learning Trajectory Generation
Carnegie Mellon University

Jun 26, 2026J/K move · Enter open · S save

Unleashing Infinite Motion: Scaling Expressive Quadrupedal Motion via Generative Video Priors

Youzhi Liu, Li Gao, Yifei Qian, Liu Liu+2

Amap, Alibaba Group

Synthetic Data Generation for Robotics Robot Motion Generation Quadruped Locomotion

Open paper AskPDF arXiv

Quadruped robots have achieved remarkable locomotion, yet their behavioral repertoire remains confined to a few gaits--far from the expressive, companion-like presence long envisioned for them. Attempts to import the humanoid recipe of large-scale motion data have inherited one tacit assumption: that robot motion must first pass through an animal body, making data collection dependent on cooperative animals, reconstruction fragile across species, and retargeting ill-posed across incompatible morphologies. We propose Uni-Mo, a fully automated pipeline that removes the animal from the loop by reframing data scarcity as a generation problem: an LLM proposes motion prompts, a video diffusion model synthesizes the corresponding robot behaviors, and the generated videos are lifted into 3D reference trajectories used to train tracking policies deployed on a real Unitree Go2. To make naively-drifting generations reliably extractable, we introduce an Identity Consistency Loss that enforces appearance coherence across frames. We release Quad-Imaginarium at https://github.com/Amap-Robotics/Quad-Imaginarium, the resulting open-source dataset of 7,488 language-annotated quadruped motions (18.5 hours) spanning acrobatic and performative behaviors. We validate 392 randomly sampled motions on a real Unitree Go2 with a 96.7% deployment success rate, complemented by a 97.6% success rate across the full dataset in simulation.

Abstract

Explore similar work

Unleashing Infinite Motion: Scaling Expressive Quadrupedal Motion via Generative Video Priors

Learning Quadruped Walking from Seconds of Demonstration

MimicAgent: Quadruped Skills via Prompt-to-Trajectory Generation