highly advanced voice recognition software written with Java scripts
Switch branches/tags
Nothing to show
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.


highly advanced voice recognition software powered by the googles and written with Java scripts

requires sox so you have to brew install sox first

also requires say so you have to have a mac

for funtimes:

npm install stenographer -g
stenographer hello yes this is dog

here is some example input and output:

input: I was seated by the shore of a small pond, about a mile and a half south of the village of Concord and somewhat higher than it

output: I was heated by the shore the small pond about a mile and a half south of the villages Concord and somewhat tired and it

you can also pipe in raw audio (use -t to specify input format):

cat hello.wav | stenographer -t wav

if the audio you pipe in is longer than 10 seconds it will automatically get chunked and processed 10 seconds at a time. googles servers return errors if you upload large files but 5-10 second clips work okay

you can control the chunk size and beginning offset using environment variables:

cat longaudio.wav | TRIMSTART=50 CHUNKSIZE=10 stenographer -t wav

if you get errors regarding input format try converting your input file to a wav first: sox input.mp3 output.wav