curl commands to use the Speech to Text service

stt.sh

#!/bin/sh
# This script clears the terminal, call the IBM Watson Speech to Text service.

USERNAME="<SERVICE_USERNAME>"
PASSWORD="<SERVICE_PASSWORD>"
SESSION_ID="<SESSION_ID>" # you will get this after running (1)
# 1. Create a session:
curl -X POST -b cookies.txt -c cookies.txt -u $USERNAME:$PASSWORD -d "{}" "https://stream.watsonplatform.net/speech-to-text/api/v1/sessions"

# This returns you a session URL.  Note that the client needs to support cookies for sessions to work.

# 2. GET as follows to fetch the interim transcription results:
curl -b cookies.txt -c cookies.txt -u $USERNAME:$PASSWORD \
"https://stream.watsonplatform.net/speech-to-text/api/v1/sessions/$SESSION_ID/observer_result?interim_results=true"

# This request will wait until the audio is submitted, and then it will return interim results in a timely manner.

# 3. POST the audio to the session recgonize URL, similar to the above examples.
# Here the audio can be sent in realtime, as it is being recorded from the system microphone.

curl -X POST -b cookies.txt -c cookies.txt -u $USERNAME:$PASSWORD \
"https://stream.watsonplatform.net/speech-to-text/api/v1/sessions/$SESSION_ID/recognize?continuous=true" --header "Content-Type: audio/flac"  --header "Transfer-Encoding: chunked" --data-binary @pcm0003.flac

# At this point you can continue submitting requests and observing interim results.  

# 4. When you are done, close the session:
curl -X POST  -b cookies.txt -c cookies.txt -u $USERNAME:$PASSWORD \
-X DELETE "https://stream.watsonplatform.net/speech-to-text/api/v1/sessions/$SESSION_ID

@germanattanasio

When I run step #1 and retrieve a session id and then set that within variable SESSION_ID, I proceed to step #2.

After executing the step, I receive the error:

curl -b cookies.txt -c cookies.txt -u $USERNAME:$PASSWORD \

"https://stream.watsonplatform.net/speech-to-text/api/v1/sessions/$SESSION_ID/observeResult?interim_results=true"

{
  "error": "Session does not exist.",
  "code": 404,
  "code_description": "No Such Resource"
}

I fixed my problem:
@germanattanasio

The curl should be updated to use observer_result and not observeResult

curl -b cookies.txt -c cookies.txt -u $USERNAME:$PASSWORD "https://stream.watsonplatform.net/speech-to-text/api/v1/sessions/$SESSION_ID/observe_result?interim_results=true"
{
"code_description": "Request Time-out",
"code": 408,
"error": "Session was closed. Reason: Session timed out due to inactivity after 30 seconds."
}

Thanks @baarongit! The api used to have observerResult and we recently updated to be observer_result. I've updated the call

line 19 comment says that audio from a live mic can be sent but the code seems to be calling for a .flac file. Is there an arecord option that writes out to a file that this monitors or can you show me how to have live audio from the mic sent in the example above?

i change line 21 and 22 to be something like...
arecord -D sysdefault:CARD=Device | curl -X POST -b cookies.txt -c cookies.txt -u $USERNAME:$PASSWORD
"https://stream.watsonplatform.net/speech-to-text/api/v1/sessions/$SESSION_ID/recognize?continuous=true" --header "Content-Type: audio/flac" --header "Transfer-Encoding: chunked" --data-binary @-

and it seems to be waiting for stdin... but it just keeps waiting nothing is returned... i suppose i am not ending the transmission so i do not ever get a response back. Is this on the right track? Can i use cURL to send live mic and get a constant stream of text back or do we have to send something, get something back and then send some more to get more back?

thank you in advance.

--jd