VoxCeleb2
PulseAugur coverage of VoxCeleb2 — every cluster mentioning VoxCeleb2 across labs, papers, and developer communities, ranked by signal.
1 day(s) with sentiment data
-
New research explores multimodal synchronization in decoder-only transformers
Researchers have explored mechanisms for improving multimodal synchronization in decoder-only transformers, specifically within video-text-to-speech (VTTS) synthesis. Their study, using the Visatronic model trained on V…
-
New SG-Mamba Model Enhances Audio-Visual Speech with Sparse Graph and Mamba
Researchers have developed SG-Mamba, a novel lightweight framework for audio-visual speech enhancement. This model integrates a sparse heterogeneous graph with a Mamba backbone to improve cross-modal alignment accuracy …
-
Language normalization method wins TidyVoice 2026 speaker verification challenge
A research paper details a method for cross-lingual speaker verification that achieved strong results in the TidyVoice 2026 Challenge. The approach involves a simple language normalization technique called Nuisance Attr…
-
Study finds mid-sized neural networks best for energy-efficient speaker verification
A new research paper evaluates the environmental impact of neural speaker verification models, focusing on energy consumption and carbon emissions during training and inference. The study analyzed ResNet architectures o…