AnyGPT: Unified Multimodal LLM with Discrete Sequence Modeling

TL;DR

AnyGPT is a versatile multimodal language model that uses discrete representations to process speech, text, images, and music seamlessly, enabling flexible multimodal interactions without altering existing LLM architectures.

Contribution

It introduces a unified discrete representation approach for multimodal processing, allowing integration of new modalities through data preprocessing without changing LLM structures.

Findings

Achieves comparable performance to specialized models across modalities

Supports any-to-any multimodal conversations

Facilitates seamless addition of new modalities

Abstract

We introduce AnyGPT, an any-to-any multimodal language model that utilizes discrete representations for the unified processing of various modalities, including speech, text, images, and music. AnyGPT can be trained stably without any alterations to the current large language model (LLM) architecture or training paradigms. Instead, it relies exclusively on data-level preprocessing, facilitating the seamless integration of new modalities into LLMs, akin to the incorporation of new languages. We build a multimodal text-centric dataset for multimodal alignment pre-training. Utilizing generative models, we synthesize the first large-scale any-to-any multimodal instruction dataset. It consists of 108k samples of multi-turn conversations that intricately interweave various modalities, thus equipping the model to handle arbitrary combinations of multimodal inputs and outputs. Experimental results demonstrate that AnyGPT is capable of facilitating any-to-any multimodal conversation while achieving performance comparable to specialized models across all modalities, proving that discrete representations can effectively and conveniently unify multiple modalities within a language model. Demos are shown in https://junzhan2000.github.io/AnyGPT.github.io/