Dia is a text-to-speech (TTS) model developed by Nari Labs, featuring 160 million parameters, capable of generating highly realistic conversations directly from text. The model supports emotion and intonation control and can generate non-verbal communication such as laughter and coughs. Its pre-trained model weights are hosted on Hugging Face and are suitable for English generation. This product is crucial for research and educational purposes, enabling advancements in conversational AI technology.

Audiblez is a tool that leverages Kokoro's high-quality speech synthesis technology to convert standard eBooks (in .epub format) into .m4b format audiobooks. It supports multiple languages and voices, allowing users to complete the conversion through simple command-line operations, greatly enriching the eBook reading experience, especially in situations where reading isn't convenient, such as while driving or exercising. This tool was developed by Claudio Santini in 2025 and is open-source under the MIT License.

CosyVoice 2 is a voice synthesis model developed by Alibaba Group's SpeechLab@Tongyi team. It is based on supervised discrete speech labels and combines two popular generative models: language models (LMs) and flow matching, achieving high naturalness, content consistency, and speaker similarity in voice synthesis. This model plays a significant role in multimodal large language models (LLMs), particularly in interactive experiences where response latency and real-time factors are crucial for speech synthesis. CosyVoice 2 enhances the utilization of speech label codebooks through limited scalar quantization, simplifies the text-to-speech language model architecture, and designs a block-aware causal flow matching model to adapt to various synthesis scenarios. It has been trained on large-scale multilingual datasets, achieving human-equivalent synthesis quality with extremely low response latency and real-time performance.

Auralis is a text-to-speech (TTS) engine that converts text into natural speech quickly, supports voice cloning, and boasts extremely fast processing speeds—capable of handling an entire novel in just minutes. The product is distinguished by its high speed, efficiency, easy integration, and high-quality audio output, making it suitable for scenarios requiring rapid text-to-speech conversion. Built on a Python API, Auralis supports long text streaming, built-in audio enhancement, automated language detection, and more. Developed by AstraMind AI, Auralis aims to provide a practical TTS solution for real-world applications. While product pricing is not explicitly stated on the page, the codebase is released under the Apache 2.0 License, allowing for free use in projects.

Fish Agent V0.1 3B is a groundbreaking speech-to-speech model capable of capturing and generating environmental audio information with unprecedented accuracy. The model utilizes a non-semantic tagging architecture, eliminating the need for traditional semantic encoders/decoders. Additionally, it is a cutting-edge text-to-speech (TTS) model trained on 700,000 hours of multilingual audio content. As a continuation of the Qwen-2.5-3B-Instruct pre-trained version, it has been trained on 200 billion speech and text tags. The model supports eight languages, including English and Chinese, with approximately 300,000 hours of training data for each of these languages and around 20,000 hours for others.

Dia is a text-to-speech (TTS) model developed by Nari Labs, featuring 160 million parameters, capable of generating highly realistic conversations directly from text. The model supports emotion and intonation control and can generate non-verbal communication such as laughter and coughs. Its pre-trained model weights are hosted on Hugging Face and are suitable for English generation. This product is crucial for research and educational purposes, enabling advancements in conversational AI technology.

Kokoro TTS is a powerful text-to-speech tool that supports multiple languages and voice blending features, capable of converting EPUB, PDF, and TXT files into high-quality speech output. The tool provides developers and users with flexible voice customization options to easily create professional-grade audio. Its main advantages include multilingual support, voice blending, flexible input formats, and a free commercial license. This product is positioned to provide creators, developers, and businesses with an efficient and low-cost speech synthesis solution, suitable for audiobook creation, video narration, podcast production, educational content generation, and customer service, among other scenarios.

Audiblez is a tool that leverages Kokoro's high-quality speech synthesis technology to convert standard eBooks (in .epub format) into .m4b format audiobooks. It supports multiple languages and voices, allowing users to complete the conversion through simple command-line operations, greatly enriching the eBook reading experience, especially in situations where reading isn't convenient, such as while driving or exercising. This tool was developed by Claudio Santini in 2025 and is open-source under the MIT License.

Flash is ElevenLabs' latest text-to-speech (TTS) model, generating speech at a speed of 75 milliseconds plus application and network latency, making it the preferred choice for low-latency, conversational voice agents. Flash v2 supports only English, while Flash v2.5 supports 32 languages, consuming 1 credit point for every two characters. In blind tests, Flash consistently outperformed other low-latency models, proving to be the fastest with guaranteed quality.

CosyVoice 2 is a voice synthesis model developed by Alibaba Group's SpeechLab@Tongyi team. It is based on supervised discrete speech labels and combines two popular generative models: language models (LMs) and flow matching, achieving high naturalness, content consistency, and speaker similarity in voice synthesis. This model plays a significant role in multimodal large language models (LLMs), particularly in interactive experiences where response latency and real-time factors are crucial for speech synthesis. CosyVoice 2 enhances the utilization of speech label codebooks through limited scalar quantization, simplifies the text-to-speech language model architecture, and designs a block-aware causal flow matching model to adapt to various synthesis scenarios. It has been trained on large-scale multilingual datasets, achieving human-equivalent synthesis quality with extremely low response latency and real-time performance.

Auralis is a text-to-speech (TTS) engine that converts text into natural speech quickly, supports voice cloning, and boasts extremely fast processing speeds—capable of handling an entire novel in just minutes. The product is distinguished by its high speed, efficiency, easy integration, and high-quality audio output, making it suitable for scenarios requiring rapid text-to-speech conversion. Built on a Python API, Auralis supports long text streaming, built-in audio enhancement, automated language detection, and more. Developed by AstraMind AI, Auralis aims to provide a practical TTS solution for real-world applications. While product pricing is not explicitly stated on the page, the codebase is released under the Apache 2.0 License, allowing for free use in projects.

OuteTTS-0.1-350M is a text-to-speech synthesis technology based on a pure language model, requiring no external adapters or complex architectures, achieving high-quality voice synthesis through carefully designed prompts and audio tokenization. This model is based on the LLaMa architecture, utilizing 350 million parameters to demonstrate the potential for direct voice synthesis using language models. It processes audio in three steps: using WavTokenizer for audio tokenization, creating precise word-to-audio mappings through CTC forced alignment, and generating structured prompts that follow specific formats. The key advantages of OuteTTS include a pure language modeling approach, voice cloning capabilities, and compatibility with llama.cpp and GGUF formats.

Fish Agent V0.1 3B is a groundbreaking speech-to-speech model capable of capturing and generating environmental audio information with unprecedented accuracy. The model utilizes a non-semantic tagging architecture, eliminating the need for traditional semantic encoders/decoders. Additionally, it is a cutting-edge text-to-speech (TTS) model trained on 700,000 hours of multilingual audio content. As a continuation of the Qwen-2.5-3B-Instruct pre-trained version, it has been trained on 200 billion speech and text tags. The model supports eight languages, including English and Chinese, with approximately 300,000 hours of training data for each of these languages and around 20,000 hours for others.