Tanel Alumäe

Parallel to the development of advanced deepfake audio generation, audio deepfake detection has also seen significant progress. However, a standardized and comprehensive benchmark is still missing. To address this, we introduce Speech DeepFake (DF) Arena, the first comprehensive benchmark for audio deepfake detection. Speech DF Arena provides a toolkit to uniformly evaluate detection systems, currently across 14 diverse datasets and attack scenarios, standardized evaluation metrics and protocols for reproducibility and transparency. It also includes a leaderboard to compare and rank the systems to help researchers and developers enhance their reliability and robustness. We include 14 evaluation sets, 14 state-of-the-art open-source and 4 proprietary detection systems, totalling 18 systems in the leaderboard. Our study presents many systems exhibiting high EER in out-of-domain scenarios, highlighting the need for extensive cross-domain evaluation. The leaderboard is hosted on HuggingFace¹ and a toolkit for reproducing results across the listed datasets is available on GitHub².

This paper describes the language identification and multilingual speech recognition system developed at Tallinn University of Technology for the Interspeech 2025 ML-SUPERB 2.0 Challenge. A hybrid language identification system is used, consisting of a pretrained language embedding model and a light-weight speech recognition model with a shared encoder across languages and language-specific bigram language models. For speech recognition, three models are used, where only a single model is applied for each language, depending on the training data availability and performance on held-out data. The model set consists of a finetuned version of SeamlessM4T, MMS-1B-all with custom language adapters and MMS-zeroshot. The system obtained the top overall score in the challenge.

This paper presents an approach for generating high-quality, same-language subtitles for Estonian TV content. We fine-tune the Whisper model on human-generated Estonian subtitles and enhance it with iterative pseudo-labeling and large language model (LLM) based post-editing. Our experiments demonstrate notable subtitle quality improvement through pseudo-labeling with an unlabeled dataset. We find that applying LLM-based editing at test time enhances subtitle accuracy, while its use during training does not yield further gains. This approach holds promise for creating subtitle quality close to human standard and could be extended to real-time applications.

This paper presents our progress in developing and maintaining a public speech and speaker recognition platform for the Estonian language. The platform consists of a speech processing pipeline and a web-based user interface for end-users, offering transcript post-editing functionality. It is offered for free as a public service and is in active use. The service provides significantly higher speech recognition accuracy than commercial alternatives. We discuss the switch to a workflow management system and how it has improved the core speech processing pipeline. The core systems behind the platform have been made available as open-source code and deployed internally by multiple public and private institutions.

In recent times, advancements in text-to-speech technologies have yielded more natural-sounding voices. However, this has also made it easier to generate malicious fake voices and disseminate false narratives. ASVspoof stands out as a prominent benchmark in the ongoing effort to automatically detect fake voices, thereby playing a crucial role in countering illicit access to biometric systems. Consequently, there is a growing need to broaden our perspectives, particularly when it comes to detecting fake voices on social media platforms. Moreover, existing detection models commonly face challenges related to their generalization performance. This study sheds light on specific instances involving the latest speech generation models. Furthermore, we introduce a novel framework designed to address the nuances of detecting fake voices in the context of social media. This framework considers not only the voice waveform but also the speech content. Our experiments have demonstrated that the proposed framework considerably enhances classification performance, as evidenced by the reduction in equal error rate. This underscores the importance of considering the waveform and the content of the voice when tasked with identifying fake voices and disseminating false claims.

This paper describes Tallinn University of Technology (TalTech) systems developed for the ASRU MADASR 2023 Challenge. The challenge focuses on automatic speech recognition of dialect-rich Indian languages with limited training audio and text data. TalTech participated in two tracks of the challenge: Track 1 that allowed using only the provided training data and Track 3 which allowed using additional audio data. In both tracks, we relied on wav2vec 2.0 models. Our methodology diverges from the traditional procedure of finetuning pretrained wav2vec 2.0 models in two key points: firstly, through the implementation of the aligned data augmentation technique to enhance the linguistic diversity of the training data, and secondly, via the application of deep prefix tuning for dialect adaptation of wav2vec 2.0 models. In both tracks, our approach yielded significant improvements over the provided baselines, achieving the lowest word error rates across all participating teams.

Language identification from speech is a common preprocessing step in many spoken language processing systems. In recent years, this field has seen fast progress, mostly due to the use of self-supervised models pretrained on multilingual data and the use of large training corpora. This paper shows that for speech with a non-native or regional accent, the accuracy of spoken language identification systems drops dramatically, and that the accuracy of identifying the language is inversely correlated with the strength of the accent. We also show that using the output of a lexicon-free speech recognition system of the particular language helps to improve language identification performance on accented speech by a large margin, without sacrificing accuracy on native speech. We obtain relative error rate reductions ranging from to 35 to 63% over the state-of-the-art model across several non-native speech datasets.