CurveCodec 2: Skeleton-agnostic animation compression with a learned entropy model

Hugging Face Daily Papers Papers
animation-compression skeleton-motion entropy-coding neural-compression rate-distortion

Summary

CurveCodec 2 introduces skeleton-agnostic skeletal animation compression that codes motion curves in the logmap with a small learned entropy model, achieving 0.37x ACL's byte usage under a worst-case per-joint error contract and 0.22x under a mean-error contract while decoding on a single CPU core and transferring without retraining to unseen species.

Skeletal motion is stored as every joint's transform at every frame, yet most of it is implied by the body rather than by what the motion is about. Compression is one way to ask what a motion must still say once the body is known, and a production codec must answer it for any skeleton with a stated error bound. Our earlier codec, CurveCodec, matched the mean error of ACL, the production library of modern game engines, with a learned prior over sparse anchors, but not ACL's worst case, and it counted its payload as floats rather than bits. Here we ask where the redundancy of skeletal motion lies and which part of a codec a learned model should take over. Measurements give three answers. At production precision the largest saving comes from predicting each quantized curve from its own past, the second from choosing per joint, in closed loop through the hierarchy, which samples not to code. On the gaps such an encoder leaves, a nearest-neighbour oracle over millions of training samples is no better than linear interpolation, and no learned in-betweener we tried paid for itself. What a network does learn is the distribution of the residuals the codec must send. CurveCodec 2 codes every sub-track as a curve in the log map, quantized in closed loop and thinned to rate-distortion-selected keys, with residuals entropy-coded under a small learned model whose integer inference is bit-exact across platforms. Two contracts are verified on every decoded clip: ACL's own worst case per joint within a stated tolerance, or ACL's mean error per clip. On a held-out test side of 4,472 clips from 33 datasets, CurveCodec 2 needs 0.37x ACL's bytes at ACL's default precision of 0.01 cm under the worst-case contract and 0.22x at 0.1 cm under the mean contract, decodes on one CPU core, and transfers without retraining to a species absent from training. Project page: https://rubbly.cn/publications/curvecodec/
Original Article
View Cached Full Text

Cached at: 10/06/26, 08:53 AM

Paper page - CurveCodec 2: Skeleton-agnostic animation compression with a learned entropy model

Source: https://huggingface.co/papers/2610.04211 Published on Oct 3

·

Submitted byhttps://huggingface.co/myshi

mingyion Oct 6

Abstract

Skeletalmotionisstoredaseveryjoint’stransformateveryframe,yetmostofitisimpliedbythebodyratherthanbywhatthemotionisabout.Compressionisonewaytoaskwhatamotionmuststillsayoncethebodyisknown,andaproductioncodecmustansweritforanyskeletonwithastatederrorbound.Ourearliercodec,CurveCodec,matchedthemeanerrorofACL,theproductionlibraryofmoderngameengines,withalearnedprioroversparseanchors,butnotACL’sworstcase,anditcounteditspayloadasfloatsratherthanbits.Hereweaskwheretheredundancyofskeletalmotionliesandwhichpartofacodecalearnedmodelshouldtakeover.Measurementsgivethreeanswers.Atproductionprecisionthelargestsavingcomesfrompredictingeachquantizedcurvefromitsownpast,thesecondfromchoosingperjoint,inclosedloopthroughthehierarchy,whichsamplesnottocode.Onthegapssuchanencoderleaves,anearest-neighbouroracleovermillionsoftrainingsamplesisnobetterthanlinearinterpolation,andnolearnedin-betweenerwetriedpaidforitself.Whatanetworkdoeslearnisthedistributionoftheresidualsthecodecmustsend.CurveCodec2codeseverysub-trackasacurveinthelogmap,quantizedinclosedloopandthinnedtorate-distortion-selectedkeys,withresidualsentropy-codedunderasmalllearnedmodelwhoseintegerinferenceisbit-exactacrossplatforms.Twocontractsareverifiedoneverydecodedclip:ACL’sownworstcaseperjointwithinastatedtolerance,orACL’smeanerrorperclip.Onaheld-outtestsideof4,472clipsfrom33datasets,CurveCodec2needs0.37xACL’sbytesatACL’sdefaultprecisionof0.01cmundertheworst-casecontractand0.22xat0.1cmunderthemeancontract,decodesononeCPUcore,andtransferswithoutretrainingtoaspeciesabsentfromtraining.Projectpage:https://rubbly.cn/publications/curvecodec/

View arXiv pageView PDFProject pageGitHubAdd to collection

Get this paper in your agent:

hf papers read 2610\.04211

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2610.04211 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2610.04211 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2610.04211 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

Data-driven Video Codec with Implicit Neural Representations

arXiv cs.AI

This paper presents a video codec that stores video and audio as weights of a sinusoidal representation network, using knowledge distillation and quantization for compression. Experiments show a 2.61x compression ratio compared to the original network, but quality lags behind standard codecs like H.264 and HEVC.

AdaCodec: A Predictive Visual Code for Video MLLMs

Hugging Face Daily Papers

AdaCodec reduces video encoding redundancy in multimodal LLMs by transmitting full visual tokens only when scene prediction fails, otherwise using compact inter-frame change descriptions. It outperforms per-frame RGB baselines at matched token budgets and achieves better or comparable results with significantly fewer tokens, reducing time-to-first-token from 9.26s to 1.62s.

Scaling Test-Time Compute for Agentic Coding

Hugging Face Daily Papers

A test-time scaling framework for agentic coding that compresses rollout trajectories into structured summaries and uses recursive voting/PDR to boost Claude-4.5-Opus to 77.6% on SWE-Bench Verified.