Fun fact, the amount of fake content grows. For example a user just sent me an article which mentions Vosk Indonesian model and even gives a link to it. The problem is we never had one! The article is clearly autogenerated!

https://x.com/FishAudio/status/1833787529595912531

Excited to introduce Fish Speech 1.4 - now open-source and more powerful than ever! Our mission is to make cutting-edge voice tech accessible to everyone.

What's new:
- Trained on 700k hours of multilingual data (up from 200k)
- Now supports 8 languages: English, Chinese, German, Japanese, French, Spanish, Korean, and Arabic
- Fully open-source, empowering developers and researchers worldwide

Key features:
- Lightning-fast TTS with ultra-low latency
- Instant voice cloning
- Self-host or use our cloud service
- Simple, flat-rate pricing

Try it out:
- Playground: https://fish.audio
- GitHub: https://github.com/fishaudio/fish-speech
- HuggingFace Model: https://huggingface.co/fishaudio/fish-speech-1.4
- Demo: https://huggingface.co/spaces/fishaudio/fish-speech-1
- Product Hunt: https://producthunt.com/posts/fish-speech-1-4

Somewhat useful, monotonic align ported to GPU

https://github.com/supertone-inc/super-monotonic-align

Small LLMS are the right thing to go, here is a Chinese LLM with 26M parameters

https://github.com/jingyaogong/minimind/blob/master/README_en.md

New paper from StyleTTS authors. Metrics looks good, and finally proper comparison between systems! But I kind of wonder if algorithms are too focused on read speech. Hard to believe in such a great metrics for conversational dataset with proposed complex algorithms.

https://arxiv.org/abs/2409.10058

StyleTTS-ZS: Efficient High-Quality Zero-Shot Text-to-Speech Synthesis with Distilled Time-Varying Style Diffusion
Yinghao Aaron Li, Xilin Jiang, Cong Han, Nima Mesgarani
The rapid development of large-scale text-to-speech (TTS) models has led to significant advancements in modeling diverse speaker prosody and voices. However, these models often face issues such as slow inference speeds, reliance on complex pre-trained neural codec representations, and difficulties in achieving naturalness and high similarity to reference speakers. To address these challenges, this work introduces StyleTTS-ZS, an efficient zero-shot TTS model that leverages distilled time-varying style diffusion to capture diverse speaker identities and prosodies. We propose a novel approach that represents human speech using input text and fixed-length time-varying discrete style codes to capture diverse prosodic variations, trained adversarially with multi-modal discriminators. A diffusion model is then built to sample this time-varying style code for efficient latent diffusion. Using classifier-free guidance, StyleTTS-ZS achieves high similarity to the reference speaker in the style diffusion process. Furthermore, to expedite sampling, the style diffusion model is distilled with perceptual loss using only 10k samples, maintaining speech quality and similarity while reducing inference speed by 90%. Our model surpasses previous state-of-the-art large-scale zero-shot TTS models in both naturalness and similarity, offering a 10-20 faster sampling speed, making it an attractive alternative for efficient large-scale zero-shot TTS systems. The audio demo, code and models are available at this https URL.

Everyone is doing DiT in TTS. Honestly I failed to understand the use of it, but metrics show good values. Examples are

https://github.com/KdaiP/StableTTS

https://github.com/7Xin/DPI-TTS

True anime dataset

https://huggingface.co/datasets/joujiboi/japanese-anime-speech-v2

Its interesting how non-optimal decisions require much more compute. For example, speech events are clearly non-uniformly arranged in time. One has to model time offsets. Given that it is strange modern discrete codecs use large uniform shifts. A proper encoder should have blank tokens or something like that and higher sampling rate.

New Whisper model large v3 turbo

https://github.com/openai/whisper/pull/2361

Linagora released Arabic Tunisian model for Vosk

https://huggingface.co/linagora/linto-asr-ar-tn-0.1

Interesting research on ASR robustness. Nemo Canary model is most accurate but far from being most Robust. Whisper variants and MMS are somewhat better. Training data size matters.

https://github.com/ahmedshah1494/speech_robust_bench

https://arxiv.org/pdf/2403.07937

Speech Robust Bench: A Robustness Benchmark For Speech Recognition

Muhammad A. Shah, David Solans Noguero, Mikko A. Heikkila, Bhiksha Raj, Nicolas Kourtellis

As Automatic Speech Recognition (ASR) models become ever more pervasive, it is important to ensure that they make reliable predictions under corruptions present in the physical and digital world. We propose Speech Robust Bench (SRB), a comprehensive benchmark for evaluating the robustness of ASR models to diverse corruptions. SRB is composed of 114 input perturbations which simulate an heterogeneous range of corruptions that ASR models may encounter when deployed in the wild. We use SRB to evaluate the robustness of several state-of-the-art ASR models and observe that model size and certain modeling choices such as the use of discrete representations, or self-training appear to be conducive to robustness. We extend this analysis to measure the robustness of ASR models on data from various demographic subgroups, namely English and Spanish speakers, and males and females. Our results revealed noticeable disparities in the model's robustness across subgroups. We believe that SRB will significantly facilitate future research towards robust ASR models, by making it easier to conduct comprehensive and comparable robustness evaluations.

https://arxiv.org/abs/2410.01036

MOSEL: 950,000 Hours of Speech Data for Open-Source Speech Foundation Model Training on EU Languages

Marco Gaido, Sara Papi, Luisa Bentivogli, Alessio Brutti, Mauro Cettolo, Roberto Gretter, Marco Matassoni, Mohamed Nabih, Matteo Negri

The rise of foundation models (FMs), coupled with regulatory efforts addressing their risks and impacts, has sparked significant interest in open-source models. However, existing speech FMs (SFMs) fall short of full compliance with the open-source principles, even if claimed otherwise, as no existing SFM has model weights, code, and training data publicly available under open-source terms. In this work, we take the first step toward filling this gap by focusing on the 24 official languages of the European Union (EU). We collect suitable training data by surveying automatic speech recognition datasets and unlabeled speech corpora under open-source compliant licenses, for a total of 950k hours. Additionally, we release automatic transcripts for 441k hours of unlabeled data under the permissive CC-BY license, thereby facilitating the creation of open-source SFMs for the EU languages.

Rev releases new very accurate English models

https://www.rev.com/blog/speech-to-text-technology/introducing-reverb-open-source-asr-diarization

Somewhat interesting

https://arxiv.org/abs/2407.15835

dMel: Speech Tokenization made Simple

He Bai, Tatiana Likhomanenko, Ruixiang Zhang, Zijin Gu, Zakaria Aldeneh, Navdeep Jaitly

Large language models have revolutionized natural language processing by leveraging self-supervised pretraining on vast textual data. Inspired by this success, researchers have investigated complicated speech tokenization methods to discretize continuous speech signals so that language modeling techniques can be applied to speech data. However, existing approaches either model semantic (content) tokens, potentially losing acoustic information, or model acoustic tokens, risking the loss of semantic (content) information. Having multiple token types also complicates the architecture and requires additional pretraining. Here we show that discretizing mel-filterbank channels into discrete intensity bins produces a simple representation (dMel), that performs better than other existing speech tokenization methods. Using an LM-style transformer architecture for speech-text modeling, we comprehensively evaluate different speech tokenization methods on speech recognition (ASR) and speech synthesis (TTS). Our results demonstrate the effectiveness of dMel in achieving high performance on both tasks within a unified framework, paving the way for efficient and effective joint modeling of speech and text.

https://arxiv.org/abs/2408.13106

NEST: Self-supervised Fast Conformer as All-purpose Seasoning to Speech Processing Tasks

He Huang, Taejin Park, Kunal Dhawan, Ivan Medennikov, Krishna C. Puvvada, Nithin Rao Koluguri, Weiqing Wang, Jagadeesh Balam, Boris Ginsburg

Self-supervised learning has been proved to benefit a wide range of speech processing tasks, such as speech recognition/translation, speaker verification and diarization, etc. However, most of current approaches are computationally expensive. In this paper, we propose a simplified and more efficient self-supervised learning framework termed as NeMo Encoder for Speech Tasks (NEST). Specifically, we adopt the FastConformer architecture with 8x sub-sampling rate, which is faster than Transformer or Conformer architectures. Instead of clustering-based quantization, we use fixed random projection for its simplicity and effectiveness. We also implement a generalized noisy speech augmentation that teaches the model to disentangle the main speaker from noise or other speakers. Experiments show that \model improves over existing self-supervised models and achieves new state-of-the-art performance on a variety of speech processing tasks, such as speech recognition/translation, speaker diarization, spoken language understanding, etc. Code and checkpoints will be publicly available via NVIDIA NeMo framework.

The more LLMs out there the more data leaks. A good example is librispeech asr accuracy of LLM is usually very good given LLM definitely saw all book texts before. Another example here

https://arxiv.org/abs/2409.04927

Just ASR + LLM? A Study on Speech Large Language Models' Ability to Identify and Understand Speaker in Spoken Dialogue

Junkai Wu, Xulin Fan, Bo-Ru Lu, Xilin Jiang, Nima Mesgarani, Mark Hasegawa-Johnson, Mari Ostendorf

In recent years, we have observed a rapid advancement in speech language models (SpeechLLMs), catching up with humans' listening and reasoning abilities. SpeechLLMs have demonstrated impressive spoken dialog question-answering (SQA) performance in benchmarks like Gaokao, the English listening test of the college entrance exam in China, which seemingly requires understanding both the spoken content and voice characteristics of speakers in a conversation. However, after carefully examining Gaokao's questions, we find the correct answers to many questions can be inferred from the conversation transcript alone, i.e.\ without speaker segmentation and identification. Our evaluation of state-of-the-art models Qwen-Audio and WavLLM on both Gaokao and our proposed "What Do You Like?" dataset shows a significantly higher accuracy in these context-based questions than in identity-critical questions, which can only be answered reliably with correct speaker identification. The results and analysis suggest that when solving SQA, the current SpeechLLMs exhibit limited speaker awareness from the audio and behave similarly to an LLM reasoning from the conversation transcription without sound. We propose that tasks focused on identity-critical questions could offer a more accurate evaluation framework of SpeechLLMs in SQA.

SLM adversarial training in styletts is not that useful (we confirmed that on our experiments too). Spectral loss is enough

Its fun how much compute resources were spent on it

Interesting that some teams still claimed it is useful, for example here in FlashSpeech https://arxiv.org/abs/2404.14700. I suppose they used only simple discriminators.

This is an important paper for training audio LLM. One can keep LM aligned with synthetic TTS data.

https://arxiv.org/abs/2309.00916

BLSP: Bootstrapping Language-Speech Pre-training via Behavior Alignment of Continuation Writing

Chen Wang, Minpeng Liao, Zhongqiang Huang, Jinliang Lu, Junhong Wu, Yuchen Liu, Chengqing Zong, Jiajun Zhang

The emergence of large language models (LLMs) has sparked significant interest in extending their remarkable language capabilities to speech. However, modality alignment between speech and text still remains an open problem. Current solutions can be categorized into two strategies. One is a cascaded approach where outputs (tokens or states) of a separately trained speech recognition system are used as inputs for LLMs, which limits their potential in modeling alignment between speech and text. The other is an end-to-end approach that relies on speech instruction data, which is very difficult to collect in large quantities. In this paper, we address these issues and propose the BLSP approach that Bootstraps Language-Speech Pre-training via behavior alignment of continuation writing. We achieve this by learning a lightweight modality adapter between a frozen speech encoder and an LLM, ensuring that the LLM exhibits the same generation behavior regardless of the modality of input: a speech segment or its transcript. The training process can be divided into two steps. The first step prompts an LLM to generate texts with speech transcripts as prefixes, obtaining text continuations. In the second step, these continuations are used as supervised signals to train the modality adapter in an end-to-end manner. We demonstrate that this straightforward process can extend the capabilities of LLMs to speech, enabling speech recognition, speech translation, spoken language understanding, and speech conversation, even in zero-shot cross-lingual scenarios.

Also

https://arxiv.org/abs/2405.19041

BLSP-KD: Bootstrapping Language-Speech Pre-training via Knowledge Distillation
Chen Wang, Minpeng Liao, Zhongqiang Huang, Jiajun Zhang
Recent end-to-end approaches have shown promise in extending large language models (LLMs) to speech inputs, but face limitations in directly assessing and optimizing alignment quality and fail to achieve fine-grained alignment due to speech-text length mismatch. We introduce BLSP-KD, a novel approach for Bootstrapping Language-Speech Pretraining via Knowledge Distillation, which addresses these limitations through two key techniques. First, it optimizes speech-text alignment by minimizing the divergence between the LLM's next-token prediction distributions for speech and text inputs using knowledge distillation. Second, it employs a continuous-integrate-andfire strategy to segment speech into tokens that correspond one-to-one with text tokens, enabling fine-grained alignment. We also introduce Partial LoRA (PLoRA), a new adaptation method supporting LLM finetuning for speech inputs under knowledge distillation. Quantitative evaluation shows that BLSP-KD outperforms previous end-to-end baselines and cascaded systems with comparable scale of parameters, facilitating general instruction-following capabilities for LLMs with speech inputs. This approach provides new possibilities for extending LLMs to spoken language interactions.