Scholarlink

SpeechCraft constitutes an innovative, modular approach towards the development of conversational AI systems. It provides a flexible framework for developing customizable voice assistants and experimenting with state-of-the-art models in speech recognition, natural language processing, and text-to-speech synthesis. The architecture is divided into three major components of the system: speech-to-text conversion, text processing for response generation using a large language model, and text-to-speech synthesis. This can be independently configured in such a way that users can choose from a range of options, including popular APIs such as OpenAI, Groq, Deepgram, and local model implementations. The key features are easy audio recording and playback, central configuration management for ease, quick prototyping, the ability to compare different models of AI, and adaptations to varied use cases; further it extends its adaptability towards language-specific models for correct processing and regional language and dialect generation. Performance evaluations reveal that the system achieves an average latency of 2.5 s, end-to-end task accuracy of 93%, and a user satisfaction score of 4.6/5, making it adaptable to diverse language and regional requirements. SpeechCraft is a valuable tool for researchers and developers to advance conversational AI systems.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

SpeechCraft: Modular AI Conversation System Using Multivariate LLMs

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Vishnu Kamisetti,
  • Goutham Bittla,
  • Dhanunjai Gujjaboina,
  • Katta Sugamya,
  • Thimmapuram Madhuri

摘要

SpeechCraft constitutes an innovative, modular approach towards the development of conversational AI systems. It provides a flexible framework for developing customizable voice assistants and experimenting with state-of-the-art models in speech recognition, natural language processing, and text-to-speech synthesis. The architecture is divided into three major components of the system: speech-to-text conversion, text processing for response generation using a large language model, and text-to-speech synthesis. This can be independently configured in such a way that users can choose from a range of options, including popular APIs such as OpenAI, Groq, Deepgram, and local model implementations. The key features are easy audio recording and playback, central configuration management for ease, quick prototyping, the ability to compare different models of AI, and adaptations to varied use cases; further it extends its adaptability towards language-specific models for correct processing and regional language and dialect generation. Performance evaluations reveal that the system achieves an average latency of 2.5 s, end-to-end task accuracy of 93%, and a user satisfaction score of 4.6/5, making it adaptable to diverse language and regional requirements. SpeechCraft is a valuable tool for researchers and developers to advance conversational AI systems.