Yanir Marmor

Yanir Marmor, Kinneret Misgav, Yair Lifshitz

We introduce "ivrit.ai", a comprehensive Hebrew speech dataset, addressing the distinct lack of extensive, high-quality resources for advancing Automated Speech Recognition (ASR) technology in Hebrew. With over 3,300 speech hours and a over a thousand diverse speakers, ivrit.ai offers a substantial compilation of Hebrew speech across various contexts. It is delivered in three forms to cater to varying research needs: raw unprocessed audio; data post-Voice Activity Detection, and partially transcribed data. The dataset stands out for its legal accessibility, permitting use at no cost, thereby serving as a crucial resource for researchers, developers, and commercial entities. ivrit.ai opens up numerous applications, offering vast potential to enhance AI capabilities in Hebrew. Future efforts aim to expand ivrit.ai further, thereby advancing Hebrew's standing in AI research and technology.

Yanir Marmor, Arad Zulti, David Krongauz et al.

Speech processing systems face a fundamental challenge: the human voice changes with age, yet few datasets support rigorous longitudinal evaluation. We introduce VoxKnesset, an open-access dataset of ~2,300 hours of Hebrew parliamentary speech spanning 2009-2025, comprising 393 speakers with recording spans of up to 15 years. Each segment includes aligned transcripts and verified demographic metadata from official parliamentary records. We benchmark modern speech embeddings (WavLM-Large, ECAPA-TDNN, Wav2Vec2-XLSR-1B) on age prediction and speaker verification under longitudinal conditions. Speaker verification EER rises from 2.15\% to 4.58\% over 15 years for the strongest model, and cross-sectionally trained age regressors fail to capture within-speaker aging, while longitudinally trained models recover a meaningful temporal signal. We publicly release the dataset and pipeline to support aging-robust speech systems and Hebrew speech processing.

Tirza Biron, Moshe Barboy, Eran Ben-Artzy et al.

Non-verbal signals in speech are encoded by prosody and carry information that ranges from conversation action to attitude and emotion. Despite its importance, the principles that govern prosodic structure are not yet adequately understood. This paper offers an analytical schema and a technological proof-of-concept for the categorization of prosodic signals and their association with meaning. The schema interprets surface-representations of multi-layered prosodic events. As a first step towards implementation, we present a classification process that disentangles prosodic phenomena of three orders. It relies on fine-tuning a pre-trained speech recognition model, enabling the simultaneous multi-class/multi-label detection. It generalizes over a large variety of spontaneous data, performing on a par with, or superior to, human annotation. In addition to a standardized formalization of prosody, disentangling prosodic patterns can direct a theory of communication and speech organization. A welcome by-product is an interpretation of prosody that will enhance speech- and language-related technologies.

Or Toledano, Yanir Marmor, Dov Gertz

Motion transfer is the task of synthesizing future video frames of a single source image according to the motion from a given driving video. In order to solve it, we face the challenging complexity of motion representation and the unknown relations between the driving video and the source image. Despite its difficulty, this problem attracted great interests from researches at the recent years, with gradual improvements. The goal is often thought as the decoupling of motion and appearance, which is may be solved by extracting the motion from keypoint movement. We chose to tackle the generic, unsupervised setting, where we need to apply animation to any arbitrary object, without any domain specific model for the structure of the input. In this work, we extract the structure from a keypoint heatmap, without an explicit motion representation. Then, the structures from the image and the video are extracted to warp the image according to the video, by a deep generator. We suggest two variants of the structure from different steps in the keypoint module, and show superior qualitative pose and quantitative scores.