Title: Mini-Omni2: Towards Open-source GPT-4o with Vision, Speech and Duplex Capabilities
Authors: Zhifei Xie, Changqiao Wu
Published: 15th October 2024 (Tuesday) @ 02:10:45
Link: http://arxiv.org/abs/2410.11190v3
Abstract
GPT-4o, an all-encompassing model, represents a milestone in the development of large multi-modal language models. It can understand visual, auditory, and textual modalities, directly output audio, and support flexible duplex interaction. Models from the open-source community often achieve some functionalities of GPT-4o, such as visual understanding and voice chat. Nevertheless, training a unified model that incorporates all modalities is challenging due to the complexities of multi-modal data, intricate model architectures, and training processes. In this paper, we introduce Mini-Omni2, a visual-audio assistant capable of providing real-time, end-to-end voice responses to visoin and audio queries. By integrating pretrained visual and auditory encoders, Mini-Omni2 maintains performance in individual modalities. We propose a three-stage training process to align modalities, allowing the language model to handle multi-modal inputs and outputs after training on a limited dataset. For interaction, we introduce a command-based interruption mechanism, enabling more flexible interaction with users. To the best of our knowledge, Mini-Omni2 is one of the closest reproductions of GPT-4o, which have similar form of functionality, and we hope it can offer valuable insights for subsequent research.
- Visual Encoder: CLIP ViT-B/32
- converts incoming images to a feature sequence of length 49 for the image patches + a global semantic feature
- concatenated to raw feature sequence of length 50
- single-layer LlamaMLP [Touvron et al., 2023] employed as the vision adapter