We present AudioNoisePrints, a training-free watermarking pipeline for flow matching and diffusion TTS models, which requires minimal extra computation during inference and does not require retraining the TTS model or reducing the generation quality. We exploited the fact that there are strong correlations between the initial Gaussian noises and the generated outputs in diffusion and flow matching models, such that a simple cosine correlation between the initial noise and the generated output can be used to perform watermaking. Moreover, we train a lightweight detector on top for more aggressive augmentations. Our method outperforms AudioSeal, a strong baseline for audio watermarking under strong augmentations. We experimented on F5TTS and other TTS and vocoder models, and concluded that they all exhibit similar spatial correlation properties, suggesting our watermarking scheme can be used for more flow-matching TTS models and even vocoders in the future.

In this article, we contrast different subjective tests for the evaluation of text-to-speech systems. We compare results obtained via MOS, a five-point absolute rating test, to two direct comparison tests: the AB test and the Best Worst Scaling (BWS) test. In an AB test, participants select their preferred stimulus from two options, while in a BWS test, they select the best and worst from multiple stimuli. We evaluate these three tests in terms of session duration, participant consistency, number of significant differences detected, and system ranking. Additionally, we estimate the impact of number of participants and test duration in their performance. To make this comparison, we conducted three separate listening tests (MOS, AB and BWS) using a common set of stimuli and analysed results derived from all collected ratings, as well as results derived from reduced subsets of listeners and stimuli. Our findings indicate that BWS tests provide more user ratings within the same amount of time. We also found that participants are more consistent with each other and themselves in the BWS test. Our results also show that the BWS test reveals more significant differences than both AB and MOS, especially with fewer participants or with shorter tests. In particular, we found that a BWS test with as few as 10 participants reveals as many differences as a MOS test with 20.

The SENCOTEN language, spoken on the Saanich peninsula of southern Vancouver Island, is in the midst of vigorous language revitalization efforts to turn the tide of language loss as a result of colonial language policies. To support these on-the-ground efforts, the community is turning to digital technology. Automatic Speech Recognition (ASR) technology holds great promise for accelerating language documentation and the creation of educational resources. However, developing ASR systems for SENCOTEN is challenging due to limited data and significant vocabulary variation from its polysynthetic structure and stress-driven metathesis. To address these challenges, we propose an ASR-driven documentation pipeline that leverages augmented speech data from a text-to-speech (TTS) system and cross-lingual transfer learning with Speech Foundation Models (SFMs). An n-gram language model is also incorporated via shallow fusion or n-best restoring to maximize the use of available data. Experiments on the SENCOTEN dataset show a word error rate (WER) of 19.34% and a character error rate (CER) of 5.09% on the test set with a 57.02% out-of-vocabulary (OOV) rate. After filtering minor cedilla-related errors, WER improves to 14.32% (26.48% on unseen words) and CER to 3.45%, demonstrating the potential of our ASR-driven pipeline to support SENCOTEN language documentation.

As the quality of contemporary speech synthesis improves, so too does the interest from language communities in developing text-to-speech (TTS) systems for a variety of real-world applications. Much of the work on TTS has focused on high-resource languages, resulting in implicitly resource-intensive paths to building such systems. The goal of this paper is to provide signposts and points of reference for future low-resource speech synthesis efforts, with insights drawn from the Speech Generation for Indigenous Language Education (SGILE) project. Funded and coordinated by the National Research Council of Canada (NRC), this multi-year, multi-partner project has the goal of producing high-quality text-to-speech systems that support the teaching of Indigenous languages in a variety of educational contexts. We provide background information and motivation for the project, as well as details about our approach and project structure, including results from a multi-day requirements-gathering session. We discuss some of our key challenges, including building models with appropriate controls for educators, improving model data efficiency, and strategies for low-resource transfer learning and evaluation. Finally, we provide a detailed survey of existing speech synthesis software and introduce EveryVoice TTS, a toolkit designed specifically for low-resource speech synthesis.

Approximate search is a valuable component of online dictionaries for learners, allowing them to find words even when they have not fully mastered the orthography or cannot reliably perceive phonemic differences in the language. However, evaluating the performance of different approximate search algorithms remains difficult in the absence of real user queries. We detail several methods for generating synthetic queries representing various user personas. We then compare the performance of several search algorithms on both real and synthetic queries in two Indigenous languages, SENCOTEN and Michif, that are phonologically and morphologically very different from English.

Despite recent progress in automatic speech recognition (ASR) and speech foundation models (SFMs) for widely spoken languages, their application to low-resource Indigenous languages remains limited. To this end, this paper presents a systematic evaluation of SFMs for ASR development in Kanyen'keha, a polysynthetic Iroquoian language structurally and typologically distinct from mainstream languages. To address challenges posed by limited data and extensive vocabulary variation, we further investigate the impact of incorporating in-domain synthesized data and external language models during cross-lingual transfer learning. Experiments on the low-resource Kanyen'keha corpus, under various train/test splits, show that the best system obtains a WER of 13.73% and a CER of 2.21% on the test set with a 59.2% OOV rate. Excluding easily correctable errors further reduces the WER and CER to 10.36% and 1.76%, demonstrating its potential to support language documentation and revitalization.

Self-supervised learning (SSL) representations from massively multilingual models offer a promising solution for low-resource language speech tasks. Despite advancements, language adaptation in TTS systems remains an open problem. This paper explores the language adaptation capability of ZMM-TTS, a recent SSL-based multilingual TTS system proposed in our previous work. We conducted experiments on 12 languages using limited data with various fine-tuning configurations. We demonstrate that the similarity in phonetics between the pre-training and target languages, as well as the language category, affects the target language's adaptation performance. Additionally, we find that the fine-tuning dataset size and number of speakers influence adaptability. Surprisingly, we also observed that using paired data for fine-tuning is not always optimal compared to audio-only data. Beyond speech intelligibility, our analysis covers speaker similarity, language identification, and predicted MOS.

Mean Opinion Score (MOS) tests are the most widely used test type for subjective evaluation of speech samples. However, their use has been questioned, as results can vary significantly depending on the test material included. Forced-choice tests such as AB or Best Worst Scaling (BWS) can in principle mitigate some of these issues. Our aim here is to compare MOS, AB and BWS tests in 3 regards: 1) Which test type do listeners prefer in terms of ease, engagement and overall likeability? 2) How fast are listeners at each test type? 3) Does each test type provide the same pattern of results? To answer these questions we re-use a subset of stimuli from the Blizzard Challenge 2013 and conduct new MOS, AB and BWS tests. Overall, we conclude each test type is broadly equally valid, MOS may not in fact be the fastest or easiest test type for listeners, but the theoretical advantages of BWS are counterbalanced by it seeming less liked by our listeners here.

Global style tokens (GSTs) allow for rich modelling of the variation in a speech corpus and subsequent control of text-to-speech synthesis (TTS). However, certain styles of speech may be marked by variation along multiple dimensions, complicating the interpretation and control of learned style tokens. One example is hyperarticulated or 'clear' speech, for example as directed toward listeners with hearing impairments or language learners in the classroom, which in English is characterised by reduced speaking rate, increased F0, more careful articulation of vowels and plosive consonants, and other factors. We present a method for simplifying control of style tokens by applying principal components analysis (PCA) to GST weights from a TTS system trained on both plain and clear speech. We identify the axes of variation in PCA space with the acoustic correlates of clear speech in English and show that we can synthesise either style by moving along a single dimension in that space. Index Terms: controllable speech synthesis, speech style

We develop an interactive web-based user interface for performing text-speech alignment and creating digital interactive "read-along" audio books that highlight words as they are spoken and allow users to replay individual words when clicked.We build on an existing Python library for zero-shot multilingual text-speech alignment (Littell et al., 2022), extend it by exposing its functionality through a RESTful API, and rewrite the underlying speech recognition engine to run in the browser.The ReadAlong Studio Web App is open-source, user-friendly, prioritizes privacy and data sovereignty, allows for a variety of standard export formats, and is designed to work for the majority of the world's languages.

While the alignment of audio recordings and text (often termed "forced alignment") is sometimes treated as a solved problem, in practice the process of adapting an alignment system to a new, under-resourced language comes with significant challenges, requiring experience and expertise that many outside of the speech community lack. This puts otherwise "solvable" problems, like the alignment of Indigenous language audiobooks, out of reach for many real-world Indigenous language organizations. In this paper, we describe ReadAlong Studio, a suite of tools for creating and visualizing aligned audiobooks, including educational features like time-aligned highlighting, playing single words in isolation, and variable-speed playback. It is intended to be accessible to creators without an extensive background in speech or NLP, by automating or making optional many of the specialist steps in an alignment pipeline. It is well documented at a beginner-technologist level, has already been adapted to 30 languages, and can work out-of-the-box on many more languages without adaptation.

This paper describes the motivation and implementation details for a rule-based, index-preserving grapheme-to-phoneme engine 'G(i)2P(i)' implemented in pure Python and released under the open source MIT license(8). The engine and interface have been designed to prioritize the developer experience of potential contributors without requiring a high level of programming knowledge. G(i)2P(i) already provides mappings for 30 (mostly Indigenous) languages, and the package is accompanied by a web-based interactive development environment, a RESTful API, and extensive documentation to encourage the addition of more mappings in the future. We also present three downstream applications of G(i)2P(i) and show results of a preliminary evaluation.

This paper describes the motivation and development of speech synthesis systems for the purposes of language revitalization. By building speech synthesis systems for three Indigenous languages spoken in Canada, Kanien’kéha, Gitksan & SENĆOŦEN, we re-evaluate the question of how much data is required to build low-resource speech synthesis systems featuring state-of-the-art neural models. For example, preliminary results with English data show that a FastSpeech2 model trained with 1 hour of training data can produce speech with comparable naturalness to a Tacotron2 model trained with 10 hours of data. Finally, we motivate future research in evaluation and classroom integration in the field of speech synthesis for language revitalization.

Roland Kuhn, Fineen Davis, Alain Désilets, Eric Joanis, Anna Kazantseva, Rebecca Knowles, Patrick Littell, Delaney Lothian, Aidan Pine, Caroline Running Wolf, Eddie Santos, Darlene Stewart, Gilles Boulianne, Vishwa Gupta, Brian Maracle Owennatékha, Akwiratékha’ Martin, Christopher Cox, Marie-Odile Junker, Olivia Sammons, Delasie Torkornoo, Nathan Thanyehténhas Brinklow, Sara Child, Benoît Farley, David Huggins-Daines, Daisy Rosenblum, Heather Souter. Proceedings of the 28th International Conference on Computational Linguistics. 2020.

Across the world’s languages and cultures, most writing systems predate the use of computers. In the early years of ICT, standards and protocols for encoding and rendering the majority of the world’s writing systems were not in place. The opportunity to deploy less-commonly used orthographies in cross-platform digital contexts has steadily increased since Unicode became the most widely used encoding on the web in late 2007 (Davis, 2008). But what happens to resources that were developed before Unicode standards became widespread? While many tools have been created to address this problem and other issues related to transliteration and character level substitutions, 1 this paper describes the process undertaken for the Indigenous and endangered Heiltsuk (Wakashan) language, and outlines a tool ( Convertextract ) that was designed to convert not only plain text, but also Microsoft Office (pptx, xlsx, docx) documents with the goals of updating and upgrading pre-existing digital textual resources to Unicode standards, and thus preserving the knowledge they contain for both the present and the future.

In this article, we discuss which text, speech, and image technologies have been developed, and would be feasible to develop, for the approximately 60 Indigenous languages spoken in Canada. In particular, we concentrate on technologies that may be feasible to develop for most or all of these languages, not just those that may be feasible for the few most-resourced of these. We assess past achievements and consider future horizons for Indigenous language transliteration, text prediction, spell-checking, approximate search, machine translation, speech recognition, speaker diarization, speech synthesis, optical character recognition, and computer-aided language learning.