1. [Publications](/publications) 2. Editing Physiological Signals in Videos Using Latent Representations # Editing Physiological Signals in Videos Using Latent Representations ![](/sites/default/files/styles/wide/public/publications/physiolatent_teaser.png?itok=jbwcXz8i) Camera-based physiological signal estimation provides a convenient and non-contact way to monitor heart rate, but it also raises serious privacy concerns because facial videos can leak sensitive information about a person’s health and emotional state. We present a learned framework for editing physiological signals in videos while preserving visual fidelity. Our method first encodes an input video into a latent representation using a pretrained 3D Variational Autoencoder, and embeds a target heart-rate prompt through a frozen text encoder. The two representations are fused by trainable spatio-temporal layers with Adaptive Layer Normalization to model the strong temporal coherence of remote photoplethysmography signals. To better preserve subtle physiological variations during reconstruction, we apply Feature-wise Linear Modulation in the decoder and fine-tune its output layer. Across multiple benchmark datasets, our approach preserves visual quality with an average PSNR of 38.96 dB and SSIM of 0.98, while achieving an average heart-rate modulation error of 10.00 bpm MAE and 10.09% MAPE under a state-of-the-art rPPG estimator. These results suggest that our framework is useful for privacy-preserving video sharing, biometric anonymization, and the generation of realistic videos with controllable vital signs. ## Authors Tianwen Zhou (University College London) Akshay Paruchuri (University of North Carolina at Chapel Hill) [Josef Spjut](/person/josef-spjut) Kaan Akşit (University College London) ## Publication Date Wednesday, June 3, 2026 ## Published in [CVPR Workshop on Subtle Visual Computing](https://sites.google.com/view/svc-cvpr26) ## Research Area [Generative AI](/research-area/generative-ai) [Human Computer Interaction](/research-area/human-computer-interaction) ## External Links [GitHub Repository](https://github.com/complight/PhysioLatent) [arxiv](https://arxiv.org/abs/2509.25348) ## Uploaded Files [Author Version](https://d1qx31qr3h6wln.cloudfront.net/publications/ZhouEtAl_CVPR2026_SVC_Workshop_Editing_physiological_signals_in_videos_using_latent_representations.pdf?VersionId=Ar5KZoOnH5ms1igbkAndFxaOQXvaWQv1 "Open file in new window")1.73 MB [Supplement](https://d1qx31qr3h6wln.cloudfront.net/publications/ZhouEtAl_CVPR2026_SVC_Workshop_Supplementary_Editing_physiological_signals_in_videos_using_latent_representations.pdf?VersionId=4vWsZeEloD0lf5pDlOoCGpIUYr9aQfAg "Open file in new window")1.59 MB ## Copyright This material is posted here with permission of the IEEE. Internal or personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution must be obtained from the IEEE by writing to .