Framework improves continuous emotion control in foundation image editors

Barycentric Guidance: Turning Foundation Image Editors into Continuous Affective Controllers

Computer Vision and Pattern RecognitionMachine Learning

Summary

Changing emotions in images is tricky because feelings aren't just on or off but continuous along two scales: how positive or negative (valence) and how energetic or calm (arousal) a face looks. Existing special-purpose tools struggle to cover all emotions smoothly and reliably, especially outside their training examples. The authors present Barycentric Guidance, a new way to steer powerful pretrained image editors to express emotions continuously without retraining them. Their method better targets specific emotional points, covers more emotional styles, and works across faces, animals, and artworks.

What this means in practice

  • For digital artists: Control facial and scene emotions smoothly in image editing without retraining models.
  • For game developers: Generate character expressions with precise emotional states continuously across different species and styles.

Authors

Harvey Mannering, Zhiwu Huang, Adam Prügel-Bennett

Abstract

Emotions are expressed on a continuum and are naturally described in the valence-arousal (VA) space. Existing VA-based image editors are typically training-based and domain-specific, limiting them to VA regions seen during training and making them brittle across datasets and domains. Meanwhile, despite producing high-quality edits, state-of-the-art foundation image editors are not continuous affective controllers: non-linear emotion representations in text encoders makes prompt interpolation poorly calibrated, so accurate and continuous VA control remains unreliable. To fill this gap, we introduce Barycentric Guidance, a training-free framework that turns pretrained foundation image editors into continuous affective controllers over VA, requiring no finetuning, extra data, or model changes. Our method maps any target VA point to simplex-based guidance weights, enabling VA controllability in three aspects: (i) accurate targeting of VA states, (ii) broad coverage across the VA plane, and (iii) continuous control. Across three face datasets, we improve VA target accuracy by 25-29%, expand VA-plane coverage by 28-50%, and increase expression diversity by 29-49%, over the strongest baseline. Unlike prior domain-specific methods, our training-free approach exploits the foundation model's inherent versatility, enabling generalization beyond human faces to animals, artworks, and complex scenes within a single framework. Code is available at https://github.com/harveymannering/BarycentricGuidance.