The article outlines a novel approach to creating a digital piano that closely mimics the sound and feel of an acoustic piano by employing machine learning and advanced sensor technology. Traditional digital pianos rely on sample-based synthesis or mathematical models, which have limitations in replicating the infinite variety of acoustic piano sounds. The authors propose a multi-modal model that integrates various sensor data, including laser, accelerometer, and MIDI sensors, to capture and analyze the performer's actions and the resulting sound. By leveraging technologies such as WaveNet for sound generation and recurrent neural networks for modeling touch data, the research aims to bridge the gap between digital and acoustic pianos. A key feature is the use of transfer learning to apply models trained on high-quality instruments to more affordable digital variants, enhancing their realism. Initial experiments have shown promising results in differentiating between digital and acoustic sounds and in predicting key dynamics, although further refinement is necessary for practical application. The work is supported by available datasets and source code, indicating a collaborative effort to advance digital musical instrument technology.