Developing a TTS system to sound like an African American voice poses challenges and reveals biases in recognition.