Back to soundboard. Goal: natural, low, dry, laconic, minimal inflection. Not robotic, not emotive. All variants avoid heavy vocoder resynthesis.
Generic neural TTS candidates approximate a dry, restrained delivery; the soundboard also has a generic browser TTS fallback. These are not Cho's actual voice and do not reproduce the actor's voice.
en-US-SteffanNeural, rate -15%, pitch -8Hz, periods only, dry EQ and light compression).
Steffan remains the selected generic TTS voice for a measured delivery, not a match to Cho or the actor. Brian and Eric are alternative male TTS candidates. Pitch-contour numbers come from tools/make_audio.py and describe delivery, not voice similarity.Only existing samples with verified male voice metadata are listed. Lessac-low was removed after the voice correction. Piper Ryan and Joe are omitted because their model cards do not specify gender; names and quality tiers are not gender labels. Browser fallback voices depend on the device and are not part of this male-only comparison.