Voice conversion (VC) is an advanced technology that enables the transformation of raw speech into high-quality audio resembling the target speaker’s voice while preserving the original linguistic content and prosodic patterns. In this study, we propose a voice conversion algorithm, Multi-Dimensional Perception Flow Matching (MPFM-VC). Unlike traditional approaches that directly generate waveform outputs, MPFM-VC models the evolutionary trajectory of mel spectrograms with a flow-matching framework and incorporates a multi-dimensional feature perception network to enhance the stability and quality of speech synthesis. Additionally, we introduce a content perturbation method during training to improve the model’s generalization ability and reduce inference-time artifacts. To further increase speaker similarity, an adversarial training mechanism on speaker embeddings is employed to achieve effective disentanglement between content and speaker identity representations, thereby enhancing the timbre consistency of the converted speech. Experimental results for both speech and singing voice conversion tasks show that MPFM-VC achieves competitive performance compared to existing state-of-the-art VC models in both subjective and objective evaluation metrics. The synthesized speech shows improved naturalness, clarity, and timbre fidelity in both objective and subjective evaluations, suggesting the potential effectiveness of the proposed approach.

Although emotional speech synthesis has seen significant progress, existing methods often struggle to generate naturally fluent emotional expression and face trade-offs between emotional richness and overall speech quality. We propose a Emotionally Expressive Speech and Singing Voice Synthesis with Large Language Models via Flow Matching, LLFM-Voice, a unified framework that enhances emotional expressiveness in both speech and singing voice synthesis. Our method leverages the contextual modeling capabilities of large language models and incorporates emotional information through an autoregressive mechanism. To further capture musical nuances in singing, we design a fine-grained emotional generator that integrates vocal techniques, tension, and pitch for precise control of expressive singing. In addition, we introduce a flow matching-based acoustic model that models the temporal evolution of mel spectrograms instead of predicting them directly, thereby mitigating artifacts introduced by conventional spectral modeling. Experiments show that LLFM-Voice outperforms baseline systems across multiple emotional expressiveness metrics, producing speech with richer emotional content and singing voices with more natural melodic expression.