TTMBA: Towards Text To Multiple Sources Binaural Audio Generation

TL;DR

This paper introduces TTMBA, a novel cascaded approach for generating multisource binaural audio from text, incorporating spatial and temporal control to enhance immersive auditory experiences.

Contribution

The paper presents a new method that combines large language models and neural networks to generate spatially accurate multisource binaural audio from text descriptions.

Findings

Outperforms existing mono text-to-audio methods in quality

Achieves accurate spatial perception in generated audio

Demonstrates effective temporal arrangement of sound sources

Abstract

Most existing text-to-audio (TTA) generation methods produce mono outputs, neglecting essential spatial information for immersive auditory experiences. To address this issue, we propose a cascaded method for text-to-multisource binaural audio generation (TTMBA) with both temporal and spatial control. First, a pretrained large language model (LLM) segments the text into a structured format with time and spatial details for each sound event. Next, a pretrained mono audio generation network creates multiple mono audios with varying durations for each event. These mono audios are transformed into binaural audios using a binaural rendering neural network based on spatial data from the LLM. Finally, the binaural audios are arranged by their start times, resulting in multisource binaural audio. Experimental results demonstrate the superiority of the proposed method in terms of both audio generation quality and spatial perceptual accuracy.