This study explores the generation of classical piano performances conditioned on individual performers' expressive styles. The interpretative diversity in classical music naturally results in comparisons and preferences for different interpretations of the same piece, fostering a desire to explore performances in varied styles. However, most research in music style transfer has primarily focused on timbre and compositional changes, with limited success in transferring performance style due to the difficulty of separating style from music content. To address this, we propose a novel approach that utilizes easily obtainable audio data and employs a flow matching model to generate high-quality, style-conditioned audio performances. We created datasets from pianists Seong-Jin Cho and Lang Lang to demonstrate our method, successfully generating performances in their distinct styles. This work expands the possibilities for performance style transfer and can potentially be applied to other performers' styles with additional audio data.