From Flat to Feeling: A Feasibility and Impact Study on Dynamic Facial Emotions in AI-Generated Avatars

Dynamic facial emotion is essential for believable AI-generated avatars; however, most systems remain visually inert, limiting their utility in high-stakes simulations such as virtual training for investigative interviews with abused children. We introduce and evaluate a real-time architecture fusing Unreal Engine 5 MetaHuman rendering with NVIDIA Omniverse Audio2Face to translate vocal prosody into high-fidelity facial expressions on photorealistic child avatars. We implemented a distributed two-PC setup that decouples language processing and speech synthesis from GPU-intensive rendering, designed to support low-latency interaction in desktop and VR environments. A between-subjects study ($N=70$) using audio+visual and visual-only conditions assessed perceptual impacts as participants rated emotional clarity, facial realism, and empathy for two avatars expressing joy, sadness, and anger. Results demonstrate that avatars could express emotions recognizably, with sadness and joy achieving high identification rates. However, anger recognition significantly dropped without audio, highlighting the importance of congruent vocal cues for high-arousal emotions. Interestingly, removing audio boosted perceived facial realism, suggesting that audiovisual desynchrony remains a key design challenge. These findings confirm the technical feasibility of generating emotionally expressive avatars and provide guidance for improving non-verbal communication in sensitive training simulations.

CMU's IWSLT 2025 Simultaneous Speech Translation System

This paper presents CMU's submission to the IWSLT 2025 Simultaneous Speech Translation (SST) task for translating unsegmented English speech into Chinese and German text in a streaming manner. Our end-to-end speech-to-text system integrates a chunkwise causal Wav2Vec 2.0 speech encoder, an adapter, and the Qwen2.5-7B-Instruct as the decoder. We use a two-stage simultaneous training procedure on robust speech segments curated from LibriSpeech, CommonVoice, and VoxPopuli datasets, utilizing standard cross-entropy loss. Our model supports adjustable latency through a configurable latency multiplier. Experimental results demonstrate that our system achieves 44.3 BLEU for English-to-Chinese and 25.1 BLEU for English-to-German translations on the ACL60/60 development set, with computation-aware latencies of 2.7 seconds and 2.3 seconds, and theoretical latencies of 2.2 and 1.7 seconds, respectively.

Instance-Specific Test-Time Training for Speech Editing in the Wild

Speech editing systems aim to naturally modify speech content while preserving acoustic consistency and speaker identity. However, previous studies often struggle to adapt to unseen and diverse acoustic conditions, resulting in degraded editing performance in real-world scenarios. To address this, we propose an instance-specific test-time training method for speech editing in the wild. Our approach employs direct supervision from ground-truth acoustic features in unedited regions, and indirect supervision in edited regions via auxiliary losses based on duration constraints and phoneme prediction. This strategy mitigates the bandwidth discontinuity problem in speech editing, ensuring smooth acoustic transitions between unedited and edited regions. Additionally, it enables precise control over speech rate by adapting the model to target durations via mask length adjustment during test-time training. Experiments on in-the-wild benchmark datasets demonstrate that our method outperforms existing speech editing systems in both objective and subjective evaluations.

SC-SOT: Conditioning the Decoder on Diarized Speaker Information for End-to-End Overlapped Speech Recognition

We propose Speaker-Conditioned Serialized Output Training (SC-SOT), an enhanced SOT-based training for E2E multi-talker ASR. We first probe how SOT handles overlapped speech, and we found the decoder performs implicit speaker separation. We hypothesize this implicit separation is often insufficient due to ambiguous acoustic cues in overlapping regions. To address this, SC-SOT explicitly conditions the decoder on speaker information, providing detailed information about "who spoke when". Specifically, we enhance the decoder by incorporating: (1) speaker embeddings, which allow the model to focus on the acoustic characteristics of the target speaker, and (2) speaker activity information, which guides the model to suppress non-target speakers. The speaker embeddings are derived from a jointly trained E2E speaker diarization model, mitigating the need for speaker enrollment. Experimental results demonstrate the effectiveness of our conditioning approach on overlapped speech.

Magnetoencephalography (MEG) Based Non-Invasive Chinese Speech Decoding

As an emerging paradigm of brain-computer interfaces (BCIs), speech BCI has the potential to directly reflect auditory perception and thoughts, offering a promising communication alternative for patients with aphasia. Chinese is one of the most widely spoken languages in the world, whereas there is very limited research on speech BCIs for Chinese language. This paper reports a text-magnetoencephalography (MEG) dataset for non-invasive Chinese speech BCIs. It also proposes a multi-modality assisted speech decoding (MASD) algorithm to capture both text and acoustic information embedded in brain signals during speech activities. Experiment results demonstrated the effectiveness of both our text-MEG dataset and our proposed MASD algorithm. To our knowledge, this is the first study on modality-assisted decoding for non-invasive speech BCIs.

Rethinking Hate Speech Detection on Social Media: Can LLMs Replace Traditional Models?

Hate speech detection across contemporary social media presents unique challenges due to linguistic diversity and the informal nature of online discourse. These challenges are further amplified in settings involving code-mixing, transliteration, and culturally nuanced expressions. While fine-tuned transformer models, such as BERT, have become standard for this task, we argue that recent large language models (LLMs) not only surpass them but also redefine the landscape of hate speech detection more broadly. To support this claim, we introduce IndoHateMix, a diverse, high-quality dataset capturing Hindi-English code-mixing and transliteration in the Indian context, providing a realistic benchmark to evaluate model robustness in complex multilingual scenarios where existing NLP methods often struggle. Our extensive experiments show that cutting-edge LLMs (such as LLaMA-3.1) consistently outperform task-specific BERT-based models, even when fine-tuned on significantly less data. With their superior generalization and adaptability, LLMs offer a transformative approach to mitigating online hate in diverse environments. This raises the question of whether future works should prioritize developing specialized models or focus on curating richer and more varied datasets to further enhance the effectiveness of LLMs.

Towards Neural Audio Codec Source Parsing

A new class of audio deepfakes-codecfakes (CFs)-has recently caught attention, synthesized by Audio Language Models that leverage neural audio codecs (NACs) in the backend. In response, the community has introduced dedicated benchmarks and tailored detection strategies. As the field advances, efforts have moved beyond binary detection toward source attribution, including open-set attribution, which aims to identify the NAC responsible for generation and flag novel, unseen ones during inference. This shift toward source attribution improves forensic interpretability and accountability. However, open-set attribution remains fundamentally limited: while it can detect that a NAC is unfamiliar, it cannot characterize or identify individual unseen codecs. It treats such inputs as generic ``unknowns'', lacking insight into their internal configuration. This leads to major shortcomings: limited generalization to new NACs and inability to resolve fine-grained variations within NAC families. To address these gaps, we propose Neural Audio Codec Source Parsing (NACSP) - a paradigm shift that reframes source attribution for CFs as structured regression over generative NAC parameters such as quantizers, bandwidth, and sampling rate. We formulate NACSP as a multi-task regression task for predicting these NAC parameters and establish the first comprehensive benchmark using various state-of-the-art speech pre-trained models (PTMs). To this end, we propose HYDRA, a novel framework that leverages hyperbolic geometry to disentangle complex latent properties from PTM representations. By employing task-specific attention over multiple curvature-aware hyperbolic subspaces, HYDRA enables superior multi-task generalization. Our extensive experiments show HYDRA achieves top results on benchmark CFs datasets compared to baselines operating in Euclidean space.

An Exploration of Mamba for Speech Self-Supervised Models

While Mamba has demonstrated strong performance in language modeling, its potential as a speech self-supervised (SSL) model remains underexplored, with prior studies limited to isolated tasks. To address this, we explore Mamba-based HuBERT models as alternatives to Transformer-based SSL architectures. Leveraging the linear-time Selective State Space, these models enable fine-tuning on long-context ASR with significantly lower compute. Moreover, they show superior performance when fine-tuned for streaming ASR. Beyond fine-tuning, these models show competitive performance on SUPERB probing benchmarks, particularly in causal settings. Our analysis shows that they yield higher-quality quantized representations and capture speaker-related features more distinctly than Transformer-based models. These findings highlight Mamba-based SSL as a promising and complementary direction for long-sequence modeling, real-time speech modeling, and speech unit extraction.

StreamMel: Real-Time Zero-shot Text-to-Speech via Interleaved Continuous Autoregressive Modeling

Recent advances in zero-shot text-to-speech (TTS) synthesis have achieved high-quality speech generation for unseen speakers, but most systems remain unsuitable for real-time applications because of their offline design. Current streaming TTS paradigms often rely on multi-stage pipelines and discrete representations, leading to increased computational cost and suboptimal system performance. In this work, we propose StreamMel, a pioneering single-stage streaming TTS framework that models continuous mel-spectrograms. By interleaving text tokens with acoustic frames, StreamMel enables low-latency, autoregressive synthesis while preserving high speaker similarity and naturalness. Experiments on LibriSpeech demonstrate that StreamMel outperforms existing streaming TTS baselines in both quality and latency. It even achieves performance comparable to offline systems while supporting efficient real-time generation, showcasing broad prospects for integration with real-time speech large language models. Audio samples are available at: https://aka.ms/StreamMel.

Mitigating Non-Target Speaker Bias in Guided Speaker Embedding

Obtaining high-quality speaker embeddings in multi-speaker conditions is crucial for many applications. A recently proposed guided speaker embedding framework, which utilizes speech activities of target and non-target speakers as clues, drastically improved embeddings under severe overlap with small degradation in low-overlap cases. However, since extreme overlaps are rare in natural conversations, this degradation cannot be overlooked. This paper first reveals that the degradation is caused by the global-statistics-based modules, widely used in speaker embedding extractors, being overly sensitive to intervals containing only non-target speakers. As a countermeasure, we propose an extension of such modules that exploit the target speaker activity clues, to compute statistics from intervals where the target is active. The proposed method improves speaker verification performance in both low and high overlap ratios, and diarization performance on multiple datasets.