Abstract Recent advancements in discrete token-based speech generation have highlighted the importance of efficient token-to-waveform synthesis in streaming and dialogue scenarios. Flow-matching acoustic decoders achieve high-quality token-to-mel generation, but their iterative sampling requires multiple neural function evaluations, limiting low-latency speech synthesis. MeanFlow reduces the sampling budget by modeling the average velocity over a temporal interval, yet maintaining high acoustic quality under extremely few-step token-to-mel generation remains challenging. To address this challenge, we propose X-Pred MeanFlow, a few-step streaming token-to-mel decoder that reparameterizes MeanFlow with mel-space prediction. The decoder predicts a generalized mel field and analytically derives the corresponding average velocity for sampling, thereby preserving the MeanFlow formulation while providing a direct acoustic prediction target. We further introduce layer-selective block-wise attention to enable continuous chunk-wise generation with bounded context. Experiments show that X-Pred MeanFlow improves few-step token-to-mel synthesis over Direct-u MeanFlow and supports stable streaming generation. Speech samples are available below.

Contents

• Model Overview
• Reconstruction Quality Comparison

This page is for research demonstration purposes only.

Model Overview

Figure 1. Overall framework and few-step inference process of the proposed X-Pred MeanFlow acoustic decoder. A frozen text-to-token LLM generates discrete speech tokens c, which condition the trainable token-to-mel decoder. At each sampling step, the decoder directly predicts the clean mel-spectrogram x̂_θ and analytically recovers the mean velocity as û_θ = (z_t − x̂_θ) / t. The resulting mel-spectrogram x̂ is converted into waveform ŷ by a frozen HiFi-GAN vocoder. The snowflake symbol denotes frozen components.