Idempotent Generative Networks (IGNs) are deep generative models that also function as local data manifold projectors, mapping arbitrary inputs back onto the manifold. They are trained to act as identity operators on the data and as idempotent operators off the data manifold. However, IGNs suffer from mode collapse, mode dropping, and training instability due to their objectives, which contain adversarial components and can cause the model to cover the data manifold only partially – an issue shared with generative adversarial networks. We introduce Non-Adversarial Idempotent Generative Networks (NAIGNs) to address these issues. Our loss function combines reconstruction with the non-adversarial generative objective of Implicit Maximum Likelihood Estimation (IMLE). This improves on IGN’s ability to restore corrupted data and generate new samples that closely match the data distribution. We moreover demonstrate that NAIGNs implicitly learn the distance field to the data manifold, as well as an energy-based model.

We demonstrate EmojiVoice, a free, customizable text-to-speech (TTS) toolkit for expressive speech on social robots. We demonstrate our voices through storytelling. This task is aimed to be deployed in classrooms, or libraries where the robot can read a story out loud to children. Moreover, we introduce adaptive clarity to to noisy environments and those with reduced comprehension ability. This storytelling robot voice allows us to demonstrate how, using our light weight and customizable TTS, we are able to have a voice that is expressive, engaging, clear and socially appropriate for the task, improving interactions with and perceptions of social robots.

Imbuing embodied agents with non-verbal behavior offers significant benefits for agent-human interactions. Despite extensive research on the creation of non-verbal behaviors, the field lacks a standardized benchmarking practice. Researchers rarely compare their findings with previous studies, and when they do, the comparisons are often not methodologically aligned. The GENEA Workshop 2025 aims to bring together the non-verbal behavior generation community to discuss major challenges and solutions in the field and determine the most effective ways to advance it.

Humans vary their expressivity when speaking for extended periods to maintain engagement with their listener. Although social robots tend to be deployed with "expressive" joyful voices, they lack this long-term variation found in human speech. Foundation model text-to-speech systems are beginning to mimic the expressivity in human speech, but they are difficult to deploy offline on robots. We present EmojiVoice, a free, customizable text-to-speech (TTS) toolkit that allows social roboticists to build temporally variable, expressive speech on social robots. We introduce emoji-prompting to allow fine-grained control of expressivity on a phase level and use the lightweight Matcha-TTS backbone to generate speech in real-time. We explore three case studies: (1) a scripted conversation with a robot assistant, (2) a storytelling robot, and (3) an autonomous speech-to-speech interactive agent. We found that using varied emoji prompting improved the perception and expressivity of speech over a long period in a storytelling task, but expressive voice was not preferred in the assistant use case.

As text-to-speech technologies achieve remarkable naturalness in read-aloud tasks, there is growing interest in multimodal synthesis of verbal and non-verbal communicative behaviour, such as spontaneous speech and associated body gestures. This paper presents a novel, unified architecture for jointly synthesising speech acoustics and skeleton-based 3D gesture motion from text, trained using optimal-transport conditional flow matching (OT-CFM). The proposed architecture is simpler than the previous state of the art, has a smaller memory footprint, and can capture the joint distribution of speech and gestures, generating both modalities together in one single process. The new training regime, meanwhile, enables better synthesis quality in much fewer steps (network evaluations) than before. Uni- and multimodal subjective tests demonstrate improved speech naturalness, gesture human-likeness, and cross-modal appropriateness compared to existing benchmarks.

We introduce Matcha-TTS, a new encoder-decoder architecture for speedy TTS acoustic modelling, trained using optimal-transport conditional flow matching (OT-CFM). This yields an ODE-based decoder capable of high output quality in fewer synthesis steps than models trained using score matching. Careful design choices additionally ensure each synthesis step is fast to run. The method is probabilistic, non-autoregressive, and learns to speak from scratch without external alignments. Compared to strong pre-trained baseline models, the Matcha-TTS system has the smallest memory footprint, rivals the speed of the fastest model on long utterances, and attains the highest mean opinion score in a listening test.

Non-verbal behavior offers significant benefits for embodied agents in human interactions. Despite extensive research on the creation of non-verbal behaviors, the field lacks a standardized benchmarking practice. Researchers rarely compare their findings with previous studies, and when they do, the comparisons are often not aligned with other methodologies. The GENEA Workshop 2024 aims to unite the community to discuss major challenges and solutions, and to determine the most effective ways to advance the field.

This article reports on the second GENEA Challenge to benchmark data-driven automatic co-speech gesture generation. Participating teams used the same speech and motion dataset to build gesture-generation systems. Motion generated by all these systems was rendered to video using a standardised visualisation pipeline and evaluated in several large, crowdsourced user studies. Unlike when comparing different research articles, differences in results are here only due to differences between methods, enabling direct comparison between systems. The dataset was based on 18 hours of full-body motion capture, including fingers, of different persons engaging in a dyadic conversation. Ten teams participated in the challenge across two tiers: full-body and upper-body gesticulation. For each tier, we evaluated both the human-likeness of the gesture motion and its appropriateness for the specific speech signal. Our evaluations decouple human-likeness from gesture appropriateness, which has been a difficult problem in the field. The evaluation results show some synthetic gesture conditions being rated as significantly more human-like than 3D human motion capture. To the best of our knowledge, this has not been demonstrated before. On the other hand, all synthetic motion is found to be vastly less appropriate for the speech than the original motion-capture recordings. We also find that conventional objective metrics do not correlate well with subjective human-likeness ratings in this large evaluation. The one exception is the Fréchet gesture distance (FGD), which achieves a Kendall’s tau rank correlation of around -0.5. Based on the challenge results we formulate numerous recommendations for system building and evaluation.

Spontaneous speech has many affective and pragmatic functions that are interesting and challenging to model in TTS. However, the presence of reduced articulation, fillers, repetitions, and other disfluencies in spontaneous speech make the text and acoustics less aligned than in read speech, which is problematic for attention-based TTS. We propose a TTS architecture that can rapidly learn to speak from small and irregular datasets, while also reproducing the diversity of expressive phenomena present in spontaneous speech. Specifically, we add utterance-level prosody control to an existing neural HMM-based TTS system which is capable of stable, monotonic alignments for spontaneous speech. We objectively evaluate control accuracy and perform perceptual tests that demonstrate that prosody control does not degrade synthesis quality. To exemplify the power of combining prosody control and ecologically valid data for reproducing intricate spontaneous speech phenomena, we evaluate the system’s capability of synthesizing two types of creaky voice.

Most state-of-the-art Text-to-Speech systems use the mel-spectrogram as an intermediate representation, to decompose the task into acoustic modelling and waveform generation.A mel-spectrogram is extracted from the waveform by a simple, fast DSP operation, but generating a high-quality waveform from a mel-spectrogram requires computationally expensive machine learning: a neural vocoder. Our proposed "autovocoder" reverses this arrangement. We use machine learning to obtain a representation that replaces the mel-spectrogram, and that can be inverted back to a waveform using simple, fast operations including a differentiable implementation of the inverse STFT.The autovocoder generates a waveform 5 times faster than the DSP-based Griffin-Lim algorithm, and 14 times faster than the neural vocoder HiFi-GAN. We provide perceptual listening test results to confirm that the speech is of comparable quality to HiFi-GAN in the copy synthesis task.

Whispering is a ubiquitous mode of communication that humans use daily. Despite this, whispered speech has been poorly served by existing speech technology due to a shortage of resources and processing methodology. To remedy this, this paper provides a processing framework that enables access to large and unique data of high-quality whispered speech. We obtain the data from recordings submitted to online platforms as part of the ASMR media-cultural phenomenon. We describe our processing pipeline and a method for improved whispered activity detection (WAD) in the ASMR data. To efficiently obtain labelled, clean whispered speech, we complement the automatic WAD by using Edyson, a bulk audio-annotation tool with human-in-the-loop. We also tackle a problem particular to ASMR: separation of whisper from other acoustic triggers present in the genre. We show that the proposed WAD and the efficient labelling allows to build extensively augmented data and train a classifier that extracts clean whisper segments from ASMR audio.Our large and growing dataset enables whisper-capable, data-driven speech technology and linguistic analysis. It also opens opportunities in e.g. HCI as a resource that may elicit emotional, psychological and neuro-physiological responses in the listener.

Neural HMMs are a type of neural transducer recently proposed for sequence-to-sequence modelling in text-to-speech.They combine the best features of classic statistical speech synthesis and modern neural TTS, requiring less data and fewer training updates, and are less prone to gibberish output caused by neural attention failures.In this paper, we combine neural HMM TTS with normalising flows for describing the highly non-Gaussian distribution of speech acoustics.The result is a powerful, fully probabilistic model of durations and acoustics that can be trained using exact maximum likelihood.Experiments show that a system based on our proposal needs fewer updates than comparable methods to produce accurate pronunciations and a subjective speech quality close to natural speech.

Abstract Gestures that accompany speech are an essential part of natural and efficient embodied human communication. The automatic generation of such co‐speech gestures is a long‐standing problem in computer animation and is considered an enabling technology for creating believable characters in film, games, and virtual social spaces, as well as for interaction with social robots. The problem is made challenging by the idiosyncratic and non‐periodic nature of human co‐speech gesture motion, and by the great diversity of communicative functions that gestures encompass. The field of gesture generation has seen surging interest in the last few years, owing to the emergence of more and larger datasets of human gesture motion, combined with strides in deep‐learning‐based generative models that benefit from the growing availability of data. This review article summarizes co‐speech gesture generation research, with a particular focus on deep generative models. First, we articulate the theory describing human gesticulation and how it complements speech. Next, we briefly discuss rule‐based and classical statistical gesture synthesis, before delving into deep learning approaches. We employ the choice of input modalities as an organizing principle, examining systems that generate gestures from audio, text and non‐linguistic input. Concurrent with the exposition of deep learning approaches, we chronicle the evolution of the related training data sets in terms of size, diversity, motion quality, and collection method (e.g., optical motion capture or pose estimation from video). Finally, we identify key research challenges in gesture generation, including data availability and quality; producing human‐like motion; grounding the gesture in the co‐occurring speech in interaction with other speakers, and in the environment; performing gesture evaluation; and integration of gesture synthesis into applications. We highlight recent approaches to tackling the various key challenges, as well as the limitations of these approaches, and point toward areas of future development.

Diffusion models have experienced a surge of interest as highly expressive yet efficiently trainable probabilistic models. We show that these models are an excellent fit for synthesising human motion that co-occurs with audio, e.g., dancing and co-speech gesticulation, since motion is complex and highly ambiguous given audio, calling for a probabilistic description. Specifically, we adapt the DiffWave architecture to model 3D pose sequences, putting Conformers in place of dilated convolutions for improved modelling power. We also demonstrate control over motion style, using classifier-free guidance to adjust the strength of the stylistic expression. Experiments on gesture and dance generation confirm that the proposed method achieves top-of-the-line motion quality, with distinctive styles whose expression can be made more or less pronounced. We also synthesise path-driven locomotion using the same model architecture. Finally, we generalise the guidance procedure to obtain product-of-expert ensembles of diffusion models and demonstrate how these may be used for, e.g., style interpolation, a contribution we believe is of independent interest.

Deep learning has revolutionised synthetic speech quality. However, it has thus far delivered little value to the speech science community. The new methods do not meet the controllability demands that practitioners in this area require e.g.: in listening tests with manipulated speech stimuli. Instead, control of different speech properties in such stimuli is achieved by using legacy signal-processing methods. This limits the range, accuracy, and speech quality of the manipulations. Also, audible artefacts have a negative impact on the methodological validity of results in speech perception studies.This work introduces a system capable of manipulating speech properties through learning rather than design. The architecture learns to control arbitrary speech properties and leverages progress in neural vocoders to obtain realistic output. Experiments with copy synthesis and manipulation of a small set of core speech features (pitch, formants, and voice quality measures) illustrate the promise of the approach for producing speech stimuli that have accurate control and high perceptual quality.

Neural sequence-to-sequence TTS has achieved significantly better output quality than statistical speech synthesis using HMMs. However, neural TTS is generally not probabilistic and uses non-monotonic attention. Attention failures increase training time and can make synthesis babble incoherently. This paper describes how the old and new paradigms can be combined to obtain the advantages of both worlds, by replacing attention in neural TTS with an autoregressive left-right no-skip hidden Markov model defined by a neural network. Based on this proposal, we modify Tacotron 2 to obtain an HMM-based neural TTS model with monotonic alignment, trained to maximise the full sequence likelihood without approximation. We also describe how to combine ideas from classical and contemporary TTS for best results. The resulting example system is smaller and simpler than Tacotron 2, and learns to speak with fewer iterations and less data, whilst achieving comparable naturalness prior to the post-net. Our approach also allows easy control over speaking rate.

Embodied agents benefit from using non-verbal behavior when communicating with humans. Despite several decades of non-verbal behavior-generation research, there is currently no well-developed benchmarking culture in the field. For example, most researchers do not compare their outcomes with previous work, and if they do, they often do so in their own way which frequently is incompatible with others. With the GENEA Workshop 2022, we aim to bring the community together to discuss key challenges and solutions, and find the most appropriate ways to move the field forward.

This paper reports on the second GENEA Challenge to benchmark data-driven automatic co-speech gesture generation. Participating teams used the same speech and motion dataset to build gesture-generation systems. Motion generated by all these systems was rendered to video using a standardised visualisation pipeline and evaluated in several large, crowdsourced user studies. Unlike when comparing different research papers, differences in results are here only due to differences between methods, enabling direct comparison between systems. This year’s dataset was based on 18 hours of full-body motion capture, including fingers, of different persons engaging in dyadic conversation. Ten teams participated in the challenge across two tiers: full-body and upper-body gesticulation. For each tier we evaluated both the human-likeness of the gesture motion and its appropriateness for the specific speech signal. Our evaluations decouple human-likeness from gesture appropriateness, which previously was a major challenge in the field.

Correct pronunciation is essential for text-to-speech (TTS) systems in production. Most production systems rely on pronouncing dictionaries to perform grapheme-to-phoneme conversion. Unlike end-to-end TTS, this enables pronunciation correction by manually altering the phoneme sequence, but the necessary dictionaries are labour-intensive to create and only exist in a few high-resourced languages. This work demonstrates that accurate TTS pronunciation control can be achieved without a dictionary. Moreover, we show that such control can be performed without requiring any model retraining or fine-tuning, merely by supplying a single correctly-pronounced reading of a word in a different voice and accent at synthesis time. Experimental results show that our proposed system successfully enables one-off correction of mispronunciations in grapheme-based TTS with maintained synthesis quality. This opens the door to production-level TTS in languages and applications where pronunciation dictionaries are unavailable.

Automatically predicting the outcome of subjective listening tests is a challenging task. Ratings may vary from person to person even if preferences are consistent across listeners. While previous work has focused on predicting listeners' ratings (mean opinion scores) of individual stimuli, we focus on the simpler task of predicting subjective preference given two speech stimuli for the same text. We propose a model based on anti-symmetric twin neural networks, trained on pairs of waveforms and their corresponding preference scores. We explore both attention and recurrent neural nets to account for the fact that stimuli in a pair are not time aligned. To obtain a large training set we convert listeners' ratings from MUSHRA tests to values that reflect how often one stimulus in the pair was rated higher than the other. Specifically, we evaluate performance on data obtained from twelve MUSHRA evaluations conducted over five years, containing different TTS systems, built from data of different speakers. Our results compare favourably to a state-of-the-art model trained to predict MOS scores.

In many research areas, for example motion and gesture generation, objective measures alone do not provide an accurate impression of key stimulus traits such as perceived quality or appropriateness. The gold standard is instead to evaluate these aspects through user studies, especially subjective evaluations of video stimuli. Common evaluation paradigms either present individual stimuli to be scored on Likert-type scales, or ask users to compare and rate videos in a pairwise fashion. However, the time and resources required for such evaluations scale poorly as the number of conditions to be compared increases. Building on standards used for evaluating the quality of multimedia codecs, this paper instead introduces a framework for granular rating of multiple comparable videos in parallel. This methodology essentially analyses all condition pairs at once. Our contributions are 1) a proposed framework, called HEMVIP, for parallel and granular evaluation of multiple video stimuli and 2) a validation study confirming that results obtained using the tool are in close agreement with results of prior studies using conventional multiple pairwise comparisons.

Embodied agents benefit from using non-verbal behavior when communicating with humans. Despite several decades of non-verbal behavior-generation research, there is currently no well-developed benchmarking culture in the field. For example, most researchers do not compare their outcomes with previous work, and if they do, they often do so in their own way which frequently is incompatible with others. With the GENEA Workshop 2021, we aim to bring the community together to discuss key challenges and solutions, and find the most appropriate ways to move the field forward.

Text-to-speech and co-speech gesture synthesis have until now been treated as separate areas by two different research communities, and applications merely stack the two technologies using a simple system-level pipeline. This can lead to modeling inefficiencies and may introduce inconsistencies that limit the achievable naturalness. We propose to instead synthesize the two modalities in a single model, a new problem we call integrated speech and gesture synthesis (ISG). We also propose a set of models modified from state-of-the-art neural speech-synthesis engines to achieve this goal. We evaluate the models in three carefully-designed user studies, two of which evaluate the synthesized speech and gesture in isolation, plus a combined study that evaluates the models like they will be used in real-world applications – speech and gesture presented together. The results show that participants rate one of the proposed integrated synthesis models as being as good as the state-of-the-art pipeline system we compare against, in all three tests. The model is able to achieve this with faster synthesis time and greatly reduced parameter count compared to the pipeline system, illustrating some of the potential benefits of treating speech and gesture synthesis together as a single, unified problem.

Co-speech gestures, gestures that accompany speech, play an important role in human communication. Automatic co-speech gesture generation is thus a key enabling technology for embodied conversational agents (ECAs), since humans expect ECAs to be capable of multi-modal communication. Research into gesture generation is rapidly gravitating towards data-driven methods. Unfortunately, individual research efforts in the field are difficult to compare: there are no established benchmarks, and each study tends to use its own dataset, motion visualisation, and evaluation methodology. To address this situation, we launched the GENEA Challenge, a gesture-generation challenge wherein participating teams built automatic gesture-generation systems on a common dataset, and the resulting systems were evaluated in parallel in a large, crowdsourced user study using the same motion-rendering pipeline. Since differences in evaluation outcomes between systems now are solely attributable to differences between the motion-generation methods, this enables benchmarking recent approaches against one another in order to get a better impression of the state of the art in the field. This paper reports on the purpose, design, results, and implications of our challenge.

We propose a new framework for gesture generation, aiming to allow data-driven approaches to produce more semantically rich gestures. Our approach first predicts whether to gesture, followed by a prediction of the gesture properties. Those properties are then used as conditioning for a modern probabilistic gesture-generation model capable of high-quality output. This empowers the approach to generate gestures that are both diverse and representational. Follow-ups and more information can be found on the project page: https://svito-zar.github.io/speech2properties2gestures/

This paper presents a novel framework for speech-driven gesture production, applicable to virtual agents to enhance human-computer interaction. Specifically, we extend recent deep-learning-based, data-driven methods for speech-driven gesture generation by incorporating representation learning. Our model takes speech as input and produces gestures as output, in the form of a sequence of 3D coordinates. We provide an analysis of different representations for the input (speech) and the output (motion) of the network by both objective and subjective evaluations. We also analyze the importance of smoothing of the produced motion. Our results indicated that the proposed method improved on our baseline in terms of objective measures. For example, it better captured the motion dynamics and better matched the motion-speed distribution. Moreover, we performed user studies on two different datasets. The studies confirmed that our proposed method is perceived as more natural than the baseline, although the difference in the studies was eliminated by appropriate post-processing: hip-centering and smoothing. We conclude that it is important to take both motion representation and post-processing into account when designing an automatic gesture-production method.

Dance requires skillful composition of complex movements that follow rhythmic, tonal and timbral features of music. Formally, generating dance conditioned on a piece of music can be expressed as a problem of modelling a high-dimensional continuous motion signal, conditioned on an audio signal. In this work we make two contributions to tackle this problem. First, we present a novel probabilistic autoregressive architecture that models the distribution over future poses with a normalizing flow conditioned on previous poses as well as music context, using a multimodal transformer encoder. Second, we introduce the currently largest 3D dance-motion dataset, obtained with a variety of motion-capture technologies, and including both professional and casual dancers. Using this dataset, we compare our new model against two baselines, via objective metrics and a user study, and show that both the ability to model a probability distribution, as well as being able to attend over a large motion and music context are necessary to produce interesting, diverse, and realistic dance that matches the music.

Breathing and speech planning in spontaneous speech are coordinated processes, often exhibiting disfluent patterns. While synthetic speech is not subject to respiratory needs, integrating breath into synthesis has advantages for naturalness and recall. At the same time, a synthetic voice reproducing disfluent breathing patterns learned from the data can be problematic. To address this, we first propose training stochastic TTS on a corpus of overlapping breath-group bigrams, to take context into account. Next, we introduce an unsupervised automatic annotation of likely-disfluent breath events, through a product-of-experts model that combines the output of two breath- event predictors, each using complementary information and operating in opposite directions. This annotation enables creating an automatically-breathing spontaneous speech synthesiser with a more fluent breathing style. A subjective evaluation on two spoken genres (impromptu and rehearsed) found the proposed system to be preferred over the baseline approach treating all breath events the same.

During speech, people spontaneously gesticulate, which plays a key role in conveying information. Similarly, realistic co-speech gestures are crucial to enable natural and smooth interactions with social agents. Current end-to-end co-speech gesture generation systems use a single modality for representing speech: either audio or text. These systems are therefore confined to producing either acoustically-linked beat gestures or semantically-linked gesticulation (e.g., raising a hand when saying "high''): they cannot appropriately learn to generate both gesture types. We present a model designed to produce arbitrary beat and semantic gestures together. Our deep-learning based model takes both acoustic and semantic representations of speech as input, and generates gestures as a sequence of joint angle rotations as output. The resulting gestures can be applied to both virtual agents and humanoid robots. Subjective and objective evaluations confirm the success of our approach. The code and video are available at the project page svito-zar.github.io/gesticulator .

Embodied human communication encompasses both verbal (speech) and non-verbal information (e.g., gesture and head movements). Recent advances in machine learning have substantially improved the technologies for generating synthetic versions of both of these types of data: On the speech side, text-to-speech systems are now able to generate highly convincing, spontaneous-sounding speech using unscripted speech audio as the source material. On the motion side, probabilistic motion-generation methods can now synthesise vivid and lifelike speech-driven 3D gesticulation. In this paper, we put these two state-of-the-art technologies together in a coherent fashion for the first time. Concretely, we demonstrate a proof-of-concept system trained on a single-speaker audio and motion-capture dataset, that is able to generate both speech and full-body gestures together from text input. In contrast to previous approaches for joint speech-and-gesture generation, we generate full-body gestures from speech synthesis trained on recordings of spontaneous speech from the same person as the motion-capture data. We illustrate our results by visualising gesture spaces and textspeech-gesture alignments, and through a demonstration video.

To enable more natural face-to-face interactions, conversational agents need to adapt their behavior to their interlocutors. One key aspect of this is generation of appropriate non-verbal behavior for the agent, for example facial gestures, here defined as facial expressions and head movements. Most existing gesture-generating systems do not utilize multi-modal cues from the interlocutor when synthesizing non-verbal behavior. Those that do, typically use deterministic methods that risk producing repetitive and non-vivid motions. In this paper, we introduce a probabilistic method to synthesize interlocutor-aware facial gestures - represented by highly expressive FLAME parameters - in dyadic conversations. Our contributions are: a) a method for feature extraction from multi-party video and speech recordings, resulting in a representation that allows for independent control and manipulation of expression and speech articulation in a 3D avatar; b) an extension to MoGlow, a recent motion-synthesis method based on normalizing flows, to also take multi-modal signals from the interlocutor as input and subsequently output interlocutor-aware facial gestures; and c) a subjective evaluation assessing the use and relative importance of the different modalities in the synthesized output. The results show that the model successfully leverages the input from the interlocutor to generate more appropriate behavior. Videos, data, and code are available at: https://jonepatr.github.io/lets_face_it/.

Abstract Automatic synthesis of realistic gestures promises to transform the fields of animation, avatars and communicative agents. In off‐line applications, novel tools can alter the role of an animator to that of a director, who provides only high‐level input for the desired animation; a learned network then translates these instructions into an appropriate sequence of body poses. In interactive scenarios, systems for generating natural animations on the fly are key to achieving believable and relatable characters. In this paper we address some of the core issues towards these ends. By adapting a deep learning‐based motion synthesis method called MoGlow, we propose a new generative model for generating state‐of‐the‐art realistic speech‐driven gesticulation. Owing to the probabilistic nature of the approach, our model can produce a battery of different, yet plausible, gestures given the same input speech signal. Just like humans, this gives a rich natural variation of motion. We additionally demonstrate the ability to exert directorial control over the output style, such as gesture level, speed, symmetry and spacial extent. Such control can be leveraged to convey a desired character personality or mood. We achieve all this without any manual annotation of the data. User studies evaluating upper‐body gesticulation confirm that the generated motions are natural and well match the input speech. Our method scores above all prior systems and baselines on these measures, and comes close to the ratings of the original recorded motions. We furthermore find that we can accurately control gesticulation styles without unnecessarily compromising perceived naturalness. Finally, we also demonstrate an application of the same method to full‐body gesticulation, including the synthesis of stepping motion and stance.

Data-driven modelling and synthesis of motion is an active research area with applications that include animation, games, and social robotics. This paper introduces a new class of probabilistic, generative, and controllable motion-data models based on normalising flows. Models of this kind can describe highly complex distributions, yet can be trained efficiently using exact maximum likelihood, unlike GANs or VAEs. Our proposed model is autoregressive and uses LSTMs to enable arbitrarily long time-dependencies. Importantly, is is also causal, meaning that each pose in the output sequence is generated without access to poses or control inputs from future time steps; this absence of algorithmic latency is important for interactive applications with real-time motion control. The approach can in principle be applied to any type of motion since it does not make restrictive, task-specific assumptions regarding the motion or the character morphology. We evaluate the models on motion-capture datasets of human and quadruped locomotion. Objective and subjective results show that randomly-sampled motion from the proposed method outperforms task-agnostic baselines and attains a motion quality close to recorded motion capture.

This paper considers utilising breaths to create improved spontaneous-speech corpora for conversational text-to-speech from found audio recordings such as dialogue podcasts. Breaths are of interest since they relate to prosody and speech planning and are independent of language and transcription. Specifically, we propose a semi-supervised approach where a fraction of coarsely annotated data is used to train a convolutional and recurrent speaker-specific breath detector operating on spectrograms and zero-crossing rate. The classifier output is used to find target-speaker breath groups (audio segments delineated by breaths) and subsequently select those that constitute clean utterances appropriate for a synthesis corpus. An application to 11 hours of raw podcast audio extracts 1969 utterances (106 minutes), 87% of which are clean and correctly segmented. This outperforms a baseline that performs integrated VAD and speaker attribution without accounting for breaths.

Synthesising spontaneous speech is a difficult task due to disfluencies, high variability and syntactic conventions different from those of written language. Using found data, as opposed to lab-rec ...

This paper presents a novel framework for automatic speech-driven gesture generation, applicable to human-agent interaction including both virtual agents and robots. Specifically, we extend recent deep-learning-based, data-driven methods for speech-driven gesture generation by incorporating representation learning. Our model takes speech as input and produces gestures as output, in the form of a sequence of 3D coordinates.

We describe a new application of deep-learning-based speech synthesis, namely multilingual speech synthesis for generating controllable foreign accent. Specifically, we train a DBLSTM-based acoustic model on non-accented multilingual speech recordings from a speaker native in several languages. By copying durations and pitch contours from a pre-recorded utterance of the desired prompt, natural prosody is achieved. We call this paradigm “cyborg speech” as it combines human and machine speech parameters. Segmentally accented speech is produced by interpolating specific quinphone linguistic features towards phones from the other language that represent non-native mispronunciations. Experiments on synthetic American-English-accented Japanese speech show that subjective synthesis quality matches monolingual synthesis, that natural pitch is maintained, and that naturalistic phone substitutions generate output that is perceived as having an American foreign accent, even though only non-accented training data was used.

Methods for adapting and controlling the characteristics of output speech are important topics in speech synthesis. In this work, we investigated the performance of DNN-based text-to-speech systems that in parallel to conventional text input also take speaker, gender, and age codes as inputs, in order to 1) perform multi-speaker synthesis, 2) perform speaker adaptation using small amounts of target-speaker adaptation data, and 3) modify synthetic speech characteristics based on the input codes. Using a large-scale, studio-quality speech corpus with 135 speakers of both genders and ages between tens and eighties, we performed three experiments: 1) First, we used a subset of speakers to construct a DNN-based, multi-speaker acoustic model with speaker codes. 2) Next, we performed speaker adaptation by estimating code vectors for new speakers via backpropagation from a small amount of adaptation material. 3) Finally, we experimented with manually manipulating input code vectors to alter the gender and/or age characteristics of the synthesised speech. Experimental results show that high-performance multi-speaker models can be constructed using the proposed code vectors with a variety of encoding schemes, and that adaptation and manipulation can be performed effectively using the codes.

For building flexible and appealing high-quality speech synthesisers, it is desirable to be able to accommodate and reproduce fine variations in vocal expression present in natural speech. Synthesisers can enable control over such output properties by adding adjustable control parameters in parallel to their text input. If not annotated in training data, the values of these control inputs can be optimised jointly with the model parameters. We describe how this established method can be seen as approximate maximum likelihood and MAP inference in a latent variable model. This puts previous ideas of (learned) synthesiser inputs such as sentence-level control vectors on a more solid theoretical footing. We furthermore extend the method by restricting the latent variables to orthogonal subspaces via a sparse prior. This enables us to learn dimensions of variation present also within classes in coarsely annotated speech. As an example, we train an LSTM-based TTS system to learn nuances in emotional expression from a speech database annotated with seven different acted emotions. Listening tests show that our proposal successfully can synthesise speech with discernible differences in expression within each emotion, without compromising the recognisability of synthesised emotions compared to an identical system without learned nuances.

This paper analyzes a) how often listeners interpret the emotional content of an utterance incorrectly when listening to vocoded or natural speech in adverse conditions; b) which noise conditions cause the most misperceptions; and c) which group of listeners misinterpret emotions the most. The long-term goal is to construct new emotional speech synthesizers that adapt to the environment and to the listener. We performed a large-scale listening test where over 400 listeners between the ages of 21 and 72 assessed natural and vocoded acted emotional speech stimuli. The stimuli had been artificially degraded using a room impulse response recorded in a car and various in-car noise types recorded in a real car. Experimental results show that the recognition rates for emotions and perceived emotional strength degrade as signal-to-noise ratio decreases. Interestingly, misperceptions seem to be more pronounced for negative and lowarousal emotions such as calmness or anger, while positive emotions such as happiness appear to be more robust to noise. An ANOVA analysis of listener meta-data further revealed that gender and age also influenced results, with elderly male listeners most likely to incorrectly identify emotions.

Forced alignment for speech synthesis traditionally aligns a phoneme sequence predetermined by the front-end text processing system. This sequence is not altered during alignment, i.e., it is forced, despite possibly being faulty. The consistency assumption is the assumption that these mistakes do not degrade models, as long as the mistakes are consistent across training and synthesis. We present evidence that in the alignment of both standard read prompts and spontaneous speech this phoneme sequence is often wrong, and that this is likely to have a negative impact on acoustic models. A lattice-based forced alignment system allowing for pronunciation variation is implemented, resulting in improved phoneme identity accuracy for both types of speech. A perceptual evaluation of HMM-based voices showed that spontaneous models trained on this improved alignment also improved standard synthesis, despite breaking the consistency assumption.

Accurate modelling and prediction of speech-sound durations is an important component in generating more natural synthetic speech. Deep neural networks (DNNs) offer a powerful modelling paradigm, and large, found corpora of natural and expressive speech are easy to acquire for training them. Unfortunately, found datasets are seldom subject to the quality-control that traditional synthesis methods expect. Common issues likely to affect duration modelling include transcription errors, reductions, filled pauses, and forced-alignment inaccuracies. To combat this, we propose to improve modelling and prediction of speech durations using methods from robust statistics, which are able to disregard ill-fitting points in the training material. We describe a robust fitting criterion based on the density power divergence (the ß-divergence) and a robust generation heuristic using mixture density networks (MDNs). Perceptual tests indicate that subjects prefer synthetic speech generated using robust models of duration over the baselines.

Deep neural networks (DNNs) have recently been the focus of much text-to-speech research as a replacement for decision trees and hidden Markov models (HMMs) in statistical parametric synthesis systems. Performance improvements have been reported; however, the configuration of systems evaluated makes it impossible to judge how much of the improvement is due to the new machine learning methods, and how much is due to other novel aspects of the systems. Specifically, whereas the decision trees in HMM-based systems typically operate at the state-level, and separate trees are used to handle separate acoustic streams, most DNN-based systems are trained to make predictions simultaneously for all streams at the level of the acoustic frame. This paper isolates the influence of three factors (machine learning method; state vs. frame predictions; separate vs. combined stream predictions) by building a continuum of systems along which only a single factor is varied at a time. We find that replacing decision trees with DNNs and moving from state-level to frame-level predictions both significantly improve listeners' naturalness ratings of synthetic speech produced by the systems. No improvement is found to result from switching from separate-stream to combined-stream predictions.

A problem when developing and tuning speech synthesis systems is that there is no well-established method of automatically rating the quality of the synthetic speech. This research attempts to obtain a new automated measure which is trained on the result of large-scale subjective evaluations employing many human listeners, i.e., the Blizzard Challenge. To exploit the data, we experiment with linear regression, feed-forward and convolutional neural network models, and combinations of them to regress from synthetic speech to the perceptual scores obtained from listeners. The biggest improvements were seen when combining stimulus- and system-level predictions.

This paper proposes a new approach to duration modelling for statistical parametric speech synthesis in which a recurrent statistical model is trained to output a phone transition probability at each timestep (acoustic frame). Unlike conventional approaches to duration modelling - which assume that duration distributions have a particular form (e.g., a Gaussian) and use the mean of that distribution for synthesis - our approach can in principle model any distribution supported on the non-negative integers. Generation from this model can be performed in many ways; here we consider output generation based on the median predicted duration. The median is more typical (more probable) than the conventional mean duration, is robust to training-data irregularities, and enables incremental generation. Furthermore, a frame-level approach to duration prediction is consistent with a longer-term goal of modelling durations and acoustic features together. Results indicate that the proposed method is competitive with baseline approaches in approximating the median duration of held-out natural speech.

We propose minimum entropy rate simplification (MERS), an information-theoretic, parameterization-independent framework for simplifying generative models of stochastic processes. Applications include improving model quality for sampling tasks by concentrating the probability mass on the most characteristic and accurately described behaviors while de-emphasizing the tails, and obtaining clean models from corrupted data (nonparametric denoising). This is the opposite of the smoothing step commonly applied to classification models. Drawing on rate-distortion theory, MERS seeks the minimum entropy-rate process under a constraint on the dissimilarity between the original and simplified processes. We particularly investigate the Kullback-Leibler divergence rate as a dissimilarity measure, where, compatible with our assumption that the starting model is disturbed or inaccurate, the simplification rather than the starting model is used for the reference distribution of the divergence. This leads to analytic solutions for stationary and ergodic Gaussian processes and Markov chains. The same formulas are also valid for maximum-entropy smoothing under the same divergence constraint. In experiments, MERS successfully simplifies and denoises models from audio, text, speech, and meteorology.

This paper presents a parametric Bayesian approach to the statistical analysis of phoneme confusion matrices measured for groups of individual listeners in one or more test conditions. Two different bias problems in conventional estimation of mutual information are analyzed and explained theoretically. Evaluations with synthetic datasets indicate that the proposed Bayesian method can give satisfactory estimates of mutual information and response probabilities, even for phoneme confusion tests using a very small number of test items for each phoneme category. The proposed method can reveal overall differences in performance between two test conditions with better power than conventional Wilcoxon significance tests or conventional confidence intervals. The method can also identify sets of confusion-matrix cells that are credibly different between two test conditions, with better power than a similar approximate frequentist method.

Tallying the numbers of listeners that took part in subjective evaluations of synthetic speech at Interspeech 2014 showed that in more than 60% of papers conclusions are based on listening tests with less than 20 listeners. Our analysis of Blizzard 2013 data shows that for a MOS test measuring naturalness a stable level of significance is only reached when more than 30 listeners are used. In this paper, we set out a list of guidelines, i.e., a checklist for carrying out meaningful subjective evaluations. We further illustrate the importance of sentence coverage and number of listeners by presenting changes to rank order and number of significant pairs by re-analysing data from the Blizzard Challenge 2013.

Parametric speech synthesis techniques depend on full context acoustic models generated by language front-ends, which anal-yse linguistic and phonetic structure. HTS, the leading paramet-ric synthesis system, can use a number of different front-ends to generate full context models for synthesis and training. In this paper we explore the use of a new text processing front-end that has been added to the speech recognition toolkit Kaldi as part of an ongoing project to produce a new parametric speech synthesis system, Idlak. The use of XML specification files, a modular design, and modern coding and testing approaches, make the Idlak front-end ideal for adding, altering and experi-menting with the contexts used in full context acoustic models. The Idlak front-end was evaluated against the standard Festival front-end in the HTS system. Results from the Idlak front-end compare well with the more mature Festival front-end (Idlak-2.83 MOS vs Festival- 2.85 MOS), although a slight reduction in naturalness perceived by non-native English speakers can be attributed to Festival’s insertion of non-punctuated pauses. Index Terms: speech synthesis, text processing, parametric synthesis, Kaldi, Idlak

Acoustic models used for statistical parametric speech synthe-sis typically incorporate many modelling assumptions. It is an open question to what extent these assumptions limit the natu-ralness of synthesised speech. To investigate this question, we recorded a speech corpus where each prompt was read aloud multiple times. By combining speech parameter trajectories ex-tracted from different repetitions, we were able to quantify the perceptual effects of certain commonly used modelling assump-tions. Subjective listening tests show that taking the source and filter parameters to be conditionally independent, or using di-agonal covariance matrices, significantly limits the naturalness that can be achieved. Our experimental results also demonstrate the shortcomings of mean-based parameter generation. Index terms: speech synthesis, acoustic modelling, stream in-dependence, diagonal covariance matrices, repeated speech 1.

An effective measure of speech intelligibility is the probability of correct recognition of the transmitted message. We propose a speech pre-enhancement method based on matching the recognized text to the text of the original message. The selected criterion is accurately approximated by the probability of the correct transcription given an estimate of the noisy speech features. In the presence of environment noise, and with a decrease in the signal-to-noise ratio, speech intelligibility declines. We implement a speech pre-enhancement system that optimizes the proposed criterion for the parameters of two distinct speech modification strategies under an energy-preservation constraint. The proposed method requires prior knowledge in the form of a transcription of the transmitted message and acoustic speech models from an automatic speech recognition system. Performance results from an open-set subjective intelligibility test indicate a significant improvement over natural speech and a reference system that optimizes a perceptual-distortion-based objective intelligibility measure. The computational complexity of the approach permits use in on-line applications.

We propose Gaussian process dynamical models (GPDMs) as a new, nonparametric paradigm in acoustic models of speech. These use multidimensional, continuous state-spaces to overcome familiar issues with discrete-state, HMM-based speech models. The added dimensions allow the state to represent and describe more than just temporal structure as systematic differences in mean, rather than as mere correlations in a residual (which dynamic features or AR-HMMs do). Being based on Gaussian processes, the models avoid restrictive parametric or linearity assumptions on signal structure. We outline GPDM theory, and describe model setup and initialization schemes relevant to speech applications. Experiments demonstrate subjectively better quality of synthesized speech than from comparable HMMs. In addition, there is evidence for unsupervised discovery of salient speech structure.

The intelligibility of speech in adverse noise conditions can be improved by modifying the characteristics of the clean speech prior to its presentation. An effective and flexible paradigm is to select the modification by optimizing a measure of objective intelligibility. Here we apply this paradigm at the text level and optimize a measure related to the classification error probability in an automatic speech recognition system. The proposed method was applied to a simple but powerful band-energy modification mechanism under an energy preservation constraint. Subjective evaluation results provide a clear indication of a significant gain in subjective intelligibility. In contrast to existing methods, the proposed approach is not restricted to a particular modification strategy and treats the notion of optimality at a level closer to that of subjective intelligibility. The computational complexity of the method is sufficiently low to enable its use in on-line applications.

Traditional discrete-state HMMs are not well suited for describing steadily evolving, path-following natural processes like motion capture data or speech. HMMs cannot represent incremental progress between behaviors, and sequences sampled from the models have unnatural segment durations, unsmooth transitions, and excessive rapid variation. We propose to address these problems by permitting the state variable to occupy positions between the discrete states, and present a concrete left-right model incorporating this idea. We call this intermediate-state HMMs. The state evolution remains Markovian. We describe training using the generalized EM-algorithm and present associated update formulas. An experiment shows that the intermediate-state model is capable of gradual transitions, with more natural durations and less noise in sampled sequences compared to a conventional HMM.