We consider the generative modeling of speech over multiple minutes, a requirement for long-form multimedia generation and audio-native voice assistants. However, textless spoken language models struggle to generate plausible speech past tens of seconds, due to high temporal resolution of speech tokens causing loss of coherence, architectural issues with long-sequence training or extrapolation, and memory costs at inference time. From these considerations we derive SpeechSSM, the first speech language model family to learn from and sample long-form spoken audio (e.g., 16 minutes of read or extemporaneous speech) in a single decoding session without text intermediates. SpeechSSMs leverage recent advances in linear-time sequence modeling to greatly surpass current Transformer spoken LMs in coherence and efficiency on multi-minute generations while still matching them at the utterance level. As we found current spoken language evaluations uninformative, especially in this new long-form setting, we also introduce: LibriSpeech-Long, a benchmark for long-form speech evaluation; new embedding-based and LLM-judged metrics; and quality measurements over length and time. Speech samples, the LibriSpeech-Long dataset, and any future code or model releases can be found at https://google.github.io/tacotron/publications/speechssm/.

In this paper, we introduce a novel Face-to-Face spoken dialogue model.It processes audio-visual speech from user input and generates audio-visual speech as the response, marking the initial step towards creating an avatar chatbot system without relying on intermediate text.To this end, we newly introduce MultiDialog, the first large-scale multimodal (i.e., audio and visual) spoken dialogue corpus containing 340 hours of approximately 9,000 dialogues, recorded based on the open domain dialogue dataset, TopicalChat.The MultiDialog contains parallel audio-visual recordings of conversation partners acting according to the given script with emotion annotations, which we expect to open up research opportunities in multimodal synthesis.Our Face-to-Face spoken dialogue model incorporates a textually pretrained large language model and adapts it into the audio-visual spoken dialogue domain by incorporating speech-text joint pretraining.Through extensive experiments, we validate the effectiveness of our model in facilitating a face-to-face conversation.

This paper proposes a novel direct Audio-Visual Speech to Audio-Visual Speech Translation (AV2AV) framework, where the input and output of the system are multimodal (i.e., audio and visual speech). With the proposed AV2AV, two key advantages can be brought: 1) We can perform real-like conversations with individuals worldwide in a virtual meeting by utilizing our own primary languages. In contrast to Speech-to-Speech Translation (A2A), which solely translates between audio modalities, the proposed AV2AV directly translates between audio-visual speech. This capability enhances the dialogue experience by presenting synchronized lip movements along with the translated speech. 2) We can improve the robustness of the spoken language translation system. By employing the complementary information of audio-visual speech, the system can effectively translate spoken language even in the presence of acoustic noise, showcasing robust performance. To mitigate the problem of the absence of a parallel AV2AV translation dataset, we propose to train our spoken language translation system with the audio-only dataset of A2A. This is done by learning unified audio-visual speech representations through self-supervised learning in advance to train the translation system. Moreover, we propose an AV-Renderer that can generate raw audio and video in parallel. It is designed with zero-shot speaker modeling, thus the speaker in source audio-visual speech can be maintained at the target translated audio-visual speech. The effectiveness of AV2AV is evaluated with extensive experiments in a many-to-many language translation setting. Demo page is available on choijeongsoo.github.io/av2av.

In this paper, we present a method for reprogramming pre-trained audio-driven talking face synthesis models to operate in a text-driven manner. Consequently, we can easily generate face videos that articulate the provided textual sentences, eliminating the necessity of recording speech for each inference, as required in the audio-driven model. To this end, we propose to embed the input text into the learned audio latent space of the pre-trained audio-driven model, while preserving the face synthesis capability of the original pretrained model. Specifically, we devise a Text-to-Audio Embedding Module (TAEM) which maps a given text input into the audio latent space by modeling pronunciation and duration characteristics. Furthermore, to consider the speaker characteristics in audio while using text inputs, TAEM is designed to accept a visual speaker embedding. The visual speaker embedding is derived from a single target face image and enables improved mapping of input text to the learned audio latent space by incorporating the speaker characteristics inherent in the audio. The main advantages of the proposed framework are that 1) it can be applied to diverse audio-driven talking face synthesis models and 2) we can generate talking face videos with either text inputs or audio inputs with high flexibility.

Providing emotional support through dialogue systems is becoming increasingly important in today’s world, as it can support both mental health and social interactions in many conversation scenarios. Previous works have shown that using persona is effective for generating empathetic and supportive responses. They have often relied on pre-provided persona rather than inferring them during conversations. However, it is not always possible to obtain a user persona before the conversation begins. To address this challenge, we propose PESS (Persona Extraction through Semantic Similarity), a novel framework that can automatically infer informative and consistent persona from dialogues. We devise completeness loss and consistency loss based on semantic similarity scores. The completeness loss encourages the model to generate missing persona information, and the consistency loss guides the model to distinguish between consistent and inconsistent persona. Our experimental results demonstrate that high-quality persona information inferred by PESS is effective in generating emotionally supportive responses.

Talking face generation is the challenging task of synthesizing a natural and realistic face that requires accurate synchronization with a given audio. Due to co-articulation, where an isolated phone is influenced by the preceding or following phones, the articulation of a phone varies upon the phonetic context. Therefore, modeling lip motion with the phonetic context can generate more spatio-temporally aligned lip movement. In this respect, we investigate the phonetic context in generating lip motion for talking face generation. We propose Context-Aware Lip-Sync framework (CALS), which explicitly leverages phonetic context to generate lip movement of the target face. CALS is comprised of an Audio-to-Lip module and a Lip-to-Face module. The former is pretrained based on masked learning to map each phone to a contextualized lip motion unit. The contextualized lip motion unit then guides the latter in synthesizing a target identity with context-aware lip motion. From extensive experiments, we verify that simply exploiting the phonetic context in the proposed CALS framework effectively enhances spatio-temporal alignment. We also demonstrate the extent to which the phonetic context assists in lip synchronization and find the effective window size for lip generation to be approximately 1.2 seconds.

This paper explores sentence-level multilingual Visual Speech Recognition (VSR) that can recognize different languages with a single trained model. As the massive multilingual modeling of visual data requires huge computational costs, we propose a novel efficient training strategy, processing with visual speech units. Through analysis, we confirm that the visual speech units mainly contain viseme information while suppressing non-linguistic information. By using the visual speech units as the inputs of our system, we propose to pre-train a VSR model to predict corresponding text outputs on multilingual data. As both the inputs (i.e., visual speech units) and outputs (i.e., text) are discrete, we can greatly improve the training efficiency compared to the standard VSR training. Specifically, the input data size is reduced to 0.016% of the original video inputs. In addition, to stabilize the training, we apply curriculum learning where the inputs of the system begin with audio-visual speech units and gradually transition to visual speech units. After pre-training, the model is finetuned on continuous features. We set new state-of-the-art multilingual VSR performances by achieving comparable performances to the previous language-specific VSR models, with a single trained model.

The challenge of talking face generation from speech lies in aligning two different modal information, audio and video, such that the mouth region corresponds to input audio. Previous methods either exploit audio-visual representation learning or leverage intermediate structural information such as landmarks and 3D models. However, they struggle to synthesize fine details of the lips varying at the phoneme level as they do not sufficiently provide visual information of the lips at the video synthesis step. To overcome this limitation, our work proposes Audio-Lip Memory that brings in visual information of the mouth region corresponding to input audio and enforces fine-grained audio-visual coherence. It stores lip motion features from sequential ground truth images in the value memory and aligns them with corresponding audio features so that they can be retrieved using audio input at inference time. Therefore, using the retrieved lip motion features as visual hints, it can easily correlate audio with visual dynamics in the synthesis step. By analyzing the memory, we demonstrate that unique lip features are stored in each memory slot at the phoneme level, capturing subtle lip motion based on memory addressing. In addition, we introduce visual-visual synchronization loss which can enhance lip-syncing performance when used along with audio-visual synchronization loss in our model. Extensive experiments are performed to verify that our method generates high-quality video with mouth shapes that best align with the input audio, outperforming previous state-of-the-art methods.

Visual Speech Recognition (VSR) is a task that recognizes speech from external appearances of the face (${\it i}.{\it e}.$, lips) into text. Since the information from the visual lip movements is not sufficient to fully represent the speech, VSR is considered as one of the challenging problems. One possible way to resolve this problem is additionally utilizing audio which contains rich information for speech recognition. However, the audio information could not be always available such as in crowded situations. Thus, it is necessary to find a way that successfully provides enough information for speech recognition with visual inputs only. In this paper, we alleviate the information insufficiency of visual lip movement by proposing a cross-modal memory augmented VSR with Visual-Audio Memory (VAM). The proposed framework tries to utilize the complementary information of audio even when the audio inputs are not provided at the inference time. Concretely, the proposed VAM learns to imprint audio features of short clip-level into a memory network using the corresponding visual features. To this end, the VAM contains two memories, lip-video key and audio value. We guide the audio value memory to imprint the audio feature and the lip-video key memory to memorize the location of the imprinted audio. By doing this, the VAM can exploit rich audio information by accessing the memory using visual inputs only. Experimental results show that the proposed method achieves state-of-the-art performance on both word- and sentence-level VSR. In addition, we verify the learned representations inside the VAM contain meaningful information for VSR.

In this paper, we introduce a novel audio-visual multi-modal bridging framework that can utilize both audio and visual information, even with uni-modal inputs. We exploit a memory network that stores source (i.e., visual) and target (i.e., audio) modal representations, where source modal representation is what we are given, and target modal representations are what we want to obtain from the memory network. We then construct an associative bridge between source and target memories that considers the inter-relationship between the two memories. By learning the interrelationship through the associative bridge, the proposed bridging framework is able to obtain the target modal representations inside the memory network, even with the source modal input only, and it provides rich information for its downstream tasks. We apply the proposed framework to two tasks: lip reading and speech reconstruction from silent video. Through the proposed associative bridge and modality-specific memories, each task knowledge is enriched with the recalled audio context, achieving state-of-the-art performance. We also verify that the associative bridge properly relates the source and target memories.

The goal of this work is to reconstruct speech from silent video, in both speaker dependent and speaker independent ways. Unlike previous works that have been mostly restricted to a speaker dependent setting, we propose Visual Voice memory to restore essential auditory information to generate proper speech from different speakers and even unseen speakers. The proposed memory takes additional auditory information that corresponds to the input face movements and stores the auditory contexts that can be recalled by the given input visual features. Specifically, the Visual Voice memory contains value and key memory slots, where value memory slots are for saving the audio features, and key memory slots are for storing the visual features in the same location of the saved audio features. Guiding each memory to properly save each feature, the model can adequately produce the speech through auxiliary information of audio. Hence, our method employs both video and audio information during training time, but does not require any additional auditory input in the inference time. Our key contributions are: (1) proposing the Visual Voice memory that brings rich information of audio that complements the visual features, thus producing high-quality speech from silent video, and (2) enabling multi-speaker and speaker independent training by memorizing auditory features and the corresponding visual features. We validate the proposed framework on GRID and Lip2Wav datasets and show that our method surpasses the performance of previous works. Moreover, we experiment on both multi-speaker and speaker independent settings and verify the effectiveness of the Visual Voice memory. We also demonstrate that the Visual Voice memory contains meaningful information to reconstruct speech.