We demonstrate Tungnaa, a software musical instrument for controlling interactive artistic text-to-voice (IATV) models. IATV is a proposed low-resource text-conditional voice generation task for musical performance applications. Tungnaá is a Python package implementing IATV training, a real-time inference engine and graphical interface for performance with IATV models. We also describe an application of Tungnaa in collaboration with sound poet Jaap Blonk.