Evaluating Idle Animation Believability: a User Perspective Publication type: Research Article Journal: Journal Volume: 00

Eneko Atxa Landa, Elena Lazkano, Igor Rodriguez, Itsaso Rodriguez-Moreno, Itziar Irigoien, Eneko Atxa Landa, Elena Lazkano, Igor Rodriguez, Itsaso Rodríguez-Moreno, Itziar Irigoien · 2025

Read Evaluating Idle Animation Believability: a User Perspective Publication type: Research Article Journal: Journal Volume: 00 by Eneko Atxa Landa, Elena Lazkano, Igor Rodriguez, Itsaso Rodriguez-Moreno, Itziar Irigoien, Eneko Atxa Landa, Elena Lazkano, Igor Rodriguez, Itsaso Rodríguez-Moreno, Itziar Irigoien (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-05023.txt.

Introduction

Avatars that merely stand still still need to breathe, shift, and look around convincingly. Eneko Atxa Landa, lead author with four co-authors, study how users judge such idle animations in a 2025 paper. Recording genuine idle movement is hard because actors who know they are recorded stop moving naturally.

Their user study reaches two conclusions. Acted and genuine idle animations are both perceived as real, and users cannot reliably tell them apart. Handmade and recorded animations are perceived differently, though. The authors argue recording datasets should become easier, and they publish a three-dimensional idle animation dataset with acted and genuine material.

Catalog record

CREATOREneko Atxa Landa, Elena Lazkano, Igor Rodriguez, Itsaso Rodriguez-Moreno, Itziar Irigoien, Eneko Atxa Landa, Elena Lazkano, Igor Rodriguez, Itsaso Rodríguez-Moreno, Itziar Irigoien
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE39 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2509-05023

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Evaluating Idle Animation Believability: a User Perspective Publication type: Research Article Journal: Journal Volume: 00 Eneko Atxa Landa, Elena Lazkano, Igor Rodriguez, Itsaso Rodriguez-Moreno, Itziar Irigoien, Eneko Atxa Landa, Elena Lazkano, Igor Rodriguez, Itsaso Rodríguez-Moreno, Itziar Irigoien Abstract Animating realistic avatars requires using high quality animations for every possible state the avatar can be in. This includes actions like walking or running, but also subtle movements that convey emotions and personality. Idle animations, such as standing, breathing or looking around, are crucial for realism and believability. In games and virtual applications, these are often handcrafted or recorded with actors, but this is costly. Furthermore, recording realistic idle animations can be very complex, because the actor must not know they are being recorded in order to make genuine movements. For this reasons idle animation datasets are not widely available. Nevertheless, this paper concludes that both acted and genuine idle animations are perceived as real, and that users are not able to distinguish between them. It also states that handmade and recorded idle animations are perceived differently. These two conclusions mean that recording idle animations should be easier than it is thought to be, meaning that actors can be specifically told to act the movements, significantly simplifying the recording process. These conclusions should help future efforts to record idle animation datasets. Finally, we also publish ReActIdle, a 3 dimensional idle animation dataset containing both real and acted idle motions. keywords Motion Capture, Idle Motion, Motion Perception, Animation † † copyright-year: 2023 † † startpage: 1 † † titlemark: Evaluating Idle Animation Believability: a User Perspective † † corresponding: Corresponding author: Eneko Atxa Landa. † † citation: . † † copyright: , , , , and . † † ctitle: Evaluating Idle Animation Believability: a User Perspective † † cjournal: Computer Animations and Virtual Worlds † † cvolume: 2021;00(00):1–18 ATXA et al. \presentaddress Computational Science and Artificial Intelligence, University of the Basque Country, Faculty of Computer Science, Donostia/ San Sebastián 1 Introduction Realism in virtual agents is affected by several aspects such as the quality of the 3D models used, the human-like behaviour of the avatar or having lifelike animations. It is also certainly an easy illusion to break: a perfectly modelled character, with a truly human-like behaviour and a realistic way of speaking can suddenly become obviously artificial if, for instance, it suddenly freezes completely after it has finished speaking. To maintain the illusion of realism, the character must have realistic behaviour at all times, with no stops or cuts in between actions. That is where idle animations come into play. These form a group of animations depicting subtle, sometimes imperceptible actions, such as breathing, looking around, making small body movements or scratching body parts. They maintain the realism of the agent, even if nothing is directly interacting with them. Such is their significance, that highly realistic virtual worlds and virtual interactive environments, especially games, implement handcrafted or pre-recorded idle animations for every character in them. However, creating these animations increases the cost of creating virtual experiences, which can pose a significant barrier for independent developers and small studios with limited budgets. In the field of motion generation, research areas such as co-speech gesture generation, 3D human motion prediction, and human motion generation are actively supported by readily available high-quality datasets. These datasets facilitate advances within these research domains. However, the availability of public good quality idle animation datasets is very scarce. One possible reason for this scarcity is the perceived complexity of capturing genuine idle movements. Firstly, recording truly natural idle behaviour requires capturing subjects “in the wild”, meaning they must be unaware of the fact that they are being observed. This is crucial as conscious awareness of being recorded would change the way they move, resulting in ungenuine movements. The recording process also presents a significant challenge: on the one hand, obtaining informed consent from individuals after the recordings is ethically problematic, and on the the other hand, once a subject has been recorded without their prior knowledge, they cannot be recorded again using the same method due to the potential psychological conditioning in the way they behave. Additionally, the use of motion capture suits is incompatible with capturing genuine idle movements. These suits generally permit recording high quality data with little noise, but their use is intrusive, because they alert the subject of the recording process and thus have an effect on the genuineness of the animations. In an effort to mitigate the insufficiency of idle animation datasets, we carried out an analysis to disprove the hypothesis that idle animation has to be recorded in a non-intrusive way. In this work, we investigate the perception of idle animations in virtual characters through user studies and variable analysis. On the one hand, we compare acted and genuine idle movements, and on the other hand, handmade and recorded movements. The process to analyse the perception of idle animations has been the following: firstly, we recorded a dataset in 3 dimensions, containing both genuine and acted idle motion. Secondly, we designed a user study in which users were shown videos containing 3D renders of genuine and acted idle animations in random order, and they were asked to classify each video in one of these two groups. The analysis of the answers has shown that users cannot correctly discriminate between genuine and acted idle animations. We have also analysed the motion data directly to compare the data distributions in terms of joint and angular speeds. The analysis has also been extended to examine the perception of handmade idle animations in comparison to the animations from the recorded dataset. 2 Related work Automatic gesture generation is a vast field of research, encompassing different tasks. In this section, an overview of some automatic gesture generation tasks and their corresponding datasets is presented. These research fields incorporate many different modalities in motion generation, such as co-speech gestures, motion prediction or text conditioned generation. The variety, quantity and quality of datasets that they have has lead to the development of generative systems in each field. Finally, some works and datasets of the idle gesture generation field are presented. 2.1 Co-speech gesture generation There are multiple papers regarding co-speech gesture generation. Qi et al. [ 1 ] propose EmotionGesture , a framework that generates 3 dimensional gestures from audio and consists of an Emotion-Beat Mining module and a Spatial Temporal Prompter module to solve the task. In another instance, Yi et al. [ 2 ] present TALKShow , generating both body and hand animations as well as face animations over a 3 dimensional mesh. In DiffGesture , [ 3 ] diffusion models are proposed to effectively capture the cross-modal audio-to-gesture association and preserve temporal coherence. Yearly, new methods for generation are presented on the Genea Challenge [ 4 ] . We refer the reader to [ 5 ] for a more comprehensive survey on co-speech gesture generation. The available datasets in this field vary in size, dimensionality, number of modalities, or whether they are monologues or dialogues, for example. Among some of the most recent datasets, BEAT [ 6 ] contains 76 hours of high-quality multimodal data of 30 speakers talking with eight different emotions and in four

Read the complete text