They had recorders and a lot of his talks are recorded, no real need to use the ai voices, there’s techniques for isolating the voice if the issue is the audience or other sounds

his talks in his own voice hit diff