We present the first demonstration of single-trial neural decoding of vowel acoustic features during speech production with high performance. The ability to predict trial-by-trial fluctuations in speech production was facilitated by using high-density, large-area electrocorticography (ECoG) combined with an adaptive principal components regression. In experiments from two human neurosurgical patients with a high-density 256-channel ECoG grid implanted over speech cortices, we demonstrate that as much as 81% of the acoustic variability across vowels could be accurately predicted from the spatial patterns of neural activity during speech production. These results demonstrate continuous, single-trial decoding of vowel acoustics.