AudioToken: Adaptation of Text-Conditioned Diffusion Models for Audio-to-Image Generation

In recent years, image generation has shown a great leap in performance, where diffusion models play a central role. Although generating high-quality images, such models are mainly conditioned on textual descriptions. This begs the question: how can we adopt such models to be conditioned on other modalities?. In this paper, we propose a novel method utilizing latent diffusion models trained for text-to-image-generation to generate images conditioned on audio recordings. Using a pre-trained audio encoding model, the proposed method encodes audio into a new token, which can be considered as an adaptation layer between the audio and text representations. Such a modeling paradigm requires a small number of trainable parameters, making the proposed approach appealing for lightweight optimization. Results suggest the proposed method is superior to the evaluated baseline methods, considering objective and subjective metrics.

We forward an audio recording through a pre-trained audio encoder and then through an Embedder network. A pre-trained text encoder extracts tokens created by a tokenizer and the audio token. Finally, the generative model is fed with the concatenated tensor of representations. It is important to note that only the Embedder (MLP+Atten-Pooling) parameters are trained during this process.