A Framework for Synthetic Audio Conversations Generation using Large Language Models
This addresses the need for synthetic audio datasets to enhance training and evaluation of audio-based AI systems like audio tagging and speech recognition, though it is incremental as it builds on existing LLM and TTS methods.
The paper tackles the problem of generating synthetic audio conversations by introducing ConversaSynth, a framework that uses large language models to create text-based dialogues and converts them to audio via text-to-speech, resulting in high-quality datasets with substantial diversity and realism.
In this paper, we introduce ConversaSynth, a framework designed to generate synthetic conversation audio using large language models (LLMs) with multiple persona settings. The framework first creates diverse and coherent text-based dialogues across various topics, which are then converted into audio using text-to-speech (TTS) systems. Our experiments demonstrate that ConversaSynth effectively generates highquality synthetic audio datasets, which can significantly enhance the training and evaluation of models for audio tagging, audio classification, and multi-speaker speech recognition. The results indicate that the synthetic datasets generated by ConversaSynth exhibit substantial diversity and realism, making them suitable for developing robust, adaptable audio-based AI systems.