Voci calcolate investigates historical and state-of-the-art speech synthesis paradigms by way of an uncanny (non-)dialogue between them and employing largely obsolete synthesis methods such as formant, concatenative, vocoder, and VOSIM, as well as more naturally sounding ones that draw from neural networks, diffusion models, and iterative denoising. As they recite text, attempt sign, or articulate vocalizations, the artificial voices are pushed beyond their limits to eventually collapse into unintentional artefacts vis-á-vis their underlying synthesis apparatuses. Keeping with, ‘voci calcolati’ celebrate semantic breakdown and technical failure are as both epistemic probes and compositional stratagems.