Turfster

i am interested in knowing the ffmpeg settings that make voices sound like Shodan. i always assumed it was some complex editing (for the skips, which is probably not the part you mean) and processing.

it's suzy @trashbang8/3/2023, 12:50 AM

In this case, I'd generated a big pile of TTS placeholder voice lines and had inadvertently saved them at a sampling rate of 16kHz. Unfortunately our engine needs them to be at 11kHz, 22kHz, or 44kHz. So I put together a script that would run 'ffmpeg -i "$file" -ar 22050 "$file" -y' on everything in the directory. This worked for most of the lines, but for some of the longer ones, the end of the line would be cut off and replaced by slowed-down versions of things the character just said.

So you'd get someone saying something like "changing the way you do busi— t h e w a y y o u d o b u s i n e s s"

JP @vectorpoem8/3/2023, 1:18 AM

ahh, that's cool. love happy accidents like that with overcomplicated tools.