MinMo: A Multimodal Large Language Model for Seamless Voice Interaction

2025-01-10

AI Summary

Current speech-text multimodal models suffer from modality processing speed mismatches, low learning efficiency, limited training data scale, and coarse-grained modeling—hindering natural, fluent full-duplex human-machine speech interaction. To address these challenges, we propose MinMo, an 8B-parameter multimodal large language model featuring a novel four-stage contrastive alignment training paradigm: speech-text, text-speech, speech-speech, and full-duplex interaction. MinMo incorporates a lightweight speech decoder and supports instruction-driven fine-grained voice control—including emotion, dialect, speaking rate, and timbre. Trained on 1.4M hours of speech data, it employs end-to-end joint modeling, instruction tuning, and a low-latency streaming inference architecture. MinMo achieves state-of-the-art performance in both speech understanding and generation. Empirical evaluation shows an end-to-end full-duplex latency of 800 ms (theoretical minimum: 600 ms) and ASR latency of ~100 ms, significantly improving instruction following and voice controllability.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Speech
• Computer Vision: Multi-modal Vision

Application Category

• Search and Retrieval-Augmented AI: Large language models for search
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Recent advancements in large language models (LLMs) and multimodal speech-text models have laid the groundwork for seamless voice interactions, enabling real-time, natural, and human-like conversations. Previous models for voice interactions are categorized as native and aligned. Native models integrate speech and text processing in one framework but struggle with issues like differing sequence lengths and insufficient pre-training. Aligned models maintain text LLM capabilities but are often limited by small datasets and a narrow focus on speech tasks. In this work, we introduce MinMo, a Multimodal Large Language Model with approximately 8B parameters for seamless voice interaction. We address the main limitations of prior aligned multimodal models. We train MinMo through multiple stages of speech-to-text alignment, text-to-speech alignment, speech-to-speech alignment, and duplex interaction alignment, on 1.4 million hours of diverse speech data and a broad range of speech tasks. After the multi-stage training, MinMo achieves state-of-the-art performance across various benchmarks for voice comprehension and generation while maintaining the capabilities of text LLMs, and also facilitates full-duplex conversation, that is, simultaneous two-way communication between the user and the system. Moreover, we propose a novel and simple voice decoder that outperforms prior models in voice generation. The enhanced instruction-following capabilities of MinMo supports controlling speech generation based on user instructions, with various nuances including emotions, dialects, and speaking rates, and mimicking specific voices. For MinMo, the speech-to-text latency is approximately 100ms, full-duplex latency is approximately 600ms in theory and 800ms in practice. The MinMo project web page is https://funaudiollm.github.io/minmo, and the code and models will be released soon.

Problem

• Asynchronous processing speed
• Low learning efficiency
• Insufficient data volume

Innovation

• MinMo Model
• Bimodal Speech-Text Processing
• Low-Latency Bidirectional Dialogue