This paper reports on the processing of generated voice for users with speech impairments, such as diglossia, harelip, and laryngectomy. The generated voice enables users to interact with the voice assistant through communication between the control interface and the voice assistant, eliminating the need for a human voice. For this purpose, speed and sound pressure level are considered. The voice is processed using a text-to-speech system in the Pyttsx3 synthesizer. It is compatible with Windows, macOS, and Linux systems. It also allows switching between male and female voices, as well as among different languages. The control system was programmed in Python. Voice assistants with noise reduction and far-field speech recognition systems can capture remote commands in noisy environments. Therefore, the limits for determining the adequate performance distance are evaluated with a 40 dB noise background, and two critical factors were identified: playback speed and sound pressure level (SPL). The system loses effectiveness if the speed is outside the range between 90 and 200 words per minute (wpm), or if the sound pressure level is below 50 dB. 100% recognition was achieved by the assistant under the following conditions: SPL ≥ 50 dB and 170 wpm.