Speech Labs is the first history of the poetry audio archive. Its aim is to define how the poetic authorial voice came to be inscribed, valued, and defined during the early period of sound recording. The central claim is that the first poetry audio archives were born in linguistic speech labs, as collaborations between poets and linguists with a mutual interest in lingual experimentation. These unlikely collaborations were precipitated by the technological complexities of making recordings during the early period of sound recording. The politics and poetics of the linguists intertwined, though did not necessarily harmonize, with those of the poets recorded, especially in relation to each’s understanding of the politics of vernacular speech and dialects. This work argues that the aesthetic compositions of each archive were shaped by scientific ways of hearing, rather than listening for entertainment or narrative absorption, and that these ways of hearing inform contemporary digital sound studies methods. This claim is furthered by archival research into the relationship of commercial record companies to recordings of poets. Arguing that record labels had a different way of hearing than the academic archivists, one informed by the profitability of vaudeville monologues, this work contends that poetry recordings have been artificially severed from apposite vocal arts by the conception that a poetry recording must bear an antecedent written text. Tracing a continuum of the aesthetics of the performative voice—from the theatricality of vaudeville-inflected poetry through the rise of the understated voice as a marker of poetic profundity—Speech Labs historicizes the voices used in poetic performance, as well as alternative conceptions of the relationship between textuality and orality. It argues that that these performances are simultaneously shaped by the technologies used for sound recording and by their archival containers—the conditions of production as poetic form.