Llasa: Scaling Train-Time and Inference-Time Compute for Llama-based Speech Synthesis

2025-02-06

AI Summary

Existing LLM-based text-to-speech (TTS) systems predominantly adopt multi-stage architectures (e.g., LLM + diffusion model), leading to complex computational scaling decisions during both training and inference. Method: We propose Llasa, an end-to-end TTS framework that pioneers *training-inference unified compute scaling* for speech synthesis. It employs a single-layer vector-quantized codec paired with a Llama-aligned Transformer architecture, enabling unified training across 1B/3B/8B model scales. Additionally, we introduce a speech understanding model as an inference-time verifier to enable verifier-guided autoregressive sampling. Contribution/Results: Llasa achieves a single-model, natively Llama-compatible, and multi-scale scalable TTS design. It significantly improves naturalness, prosodic complexity, emotional expressiveness, timbre consistency, and content fidelity. All models and code are open-sourced.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Computer Vision: Large Vision Models

Application Category

• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

Recent advances in text-based large language models (LLMs), particularly in the GPT series and the o1 model, have demonstrated the effectiveness of scaling both training-time and inference-time compute. However, current state-of-the-art TTS systems leveraging LLMs are often multi-stage, requiring separate models (e.g., diffusion models after LLM), complicating the decision of whether to scale a particular model during training or testing. This work makes the following contributions: First, we explore the scaling of train-time and inference-time compute for speech synthesis. Second, we propose a simple framework Llasa for speech synthesis that employs a single-layer vector quantizer (VQ) codec and a single Transformer architecture to fully align with standard LLMs such as Llama. Our experiments reveal that scaling train-time compute for Llasa consistently improves the naturalness of synthesized speech and enables the generation of more complex and accurate prosody patterns. Furthermore, from the perspective of scaling inference-time compute, we employ speech understanding models as verifiers during the search, finding that scaling inference-time compute shifts the sampling modes toward the preferences of specific verifiers, thereby improving emotional expressiveness, timbre consistency, and content accuracy. In addition, we released the checkpoint and training code for our TTS model (1B, 3B, 8B) and codec model publicly available.

Problem

• Scaling train-time and inference-time compute
• Simplified framework for speech synthesis
• Improving naturalness and prosody patterns

Innovation

• Single-layer VQ codec
• Transformer architecture alignment
• Scaling compute for speech synthesis