Quick and dirty python implementation of client for google speech to text service using google speech stream API. It requires requests, pyaudio and pyflac to be installed. Don't forget to set your API key. Current limit seems to be around 5 minutes.

google_stt.py

#!/usr/bin/python

import flac.encoder as encoder
import pyaudio
import sys
import requests
import random

from threading import Thread
from Queue import Queue, Empty
from time import sleep

class google_stt_stream(object):
    def __init__(self):
        self.write_queue = Queue()
        self.keep_streaming = True

        self.upstream_url = "https://www.google.com/speech-api/full-duplex/v1/up?key=%(key)s&pair=%(pair)s&lang=en-US&maxAlternatives=20&client=chromium&continuous&interim"
        self.upstream_headers = {'content-type': 'audio/x-flac; rate=16000'}
        self.downstream_url = "https://www.google.com/speech-api/full-duplex/v1/down?pair=%(pair)s"
        self.api_key = "<your_api_key>"

    def generate_request_key(self):
        return hex(random.getrandbits(64))[2:-1]

    def start(self):
        pair = self.generate_request_key()
        upstream_url = self.upstream_url % {"pair": pair, "key": self.api_key}
        downstream_url = self.downstream_url % {"pair": pair, "key": self.api_key}

        self.session = requests.Session()
        self.upstream_thread = Thread(target=self.upstream, args=(upstream_url,))
        self.downstream_thread = Thread(target=self.downstream, args=(downstream_url,))

        self.downstream_thread.start()
        self.upstream_thread.start()
 
    def stop(self):
        print "Waiting write_queue to write all data"
        self.write_queue.join()
        print "Queue empty"
        sleep(10)

        self.keep_streaming=False
        self.upstream_thread.join()
        self.downstream_thread.join()

    def write_data(self, data):
        self.write_queue.put(data)

    def gen_data(self):
        while self.keep_streaming:
            try:
                item = self.write_queue.get(timeout=2)
            except Empty:
                return
            yield item
            self.write_queue.task_done()

    def upstream(self, url):
        print self.session.post(url, headers=self.upstream_headers, data=self.gen_data())

    def downstream(self, url):
        r = self.session.get(url, stream=True)
        while self.keep_streaming:
            try:
                for line in r.iter_content():
                    if not self.keep_streaming:
                        break
                    if line:
                        sys.stdout.write(line)
            except Exception as e:
                print "Exception %s, restarting" %e
                self.keep_streaming = False
                self.upstream_thread.join()
                self.keep_streaming = True
                self.start()
                return

        print "end"

stt = google_stt_stream()

def write(enc, buf, samples, current_frame):
    stt.write_data(buf)
    #print current_frame, samples
    return True

#config
chunk = 512
FORMAT = pyaudio.paInt16
CHANNELS = 1
RATE = 16000
THRESHOLD = 180 #The threshold intensity that defines silence signal (lower than).

#open stream
p = pyaudio.PyAudio()

stream = p.open(format = FORMAT,
                channels = CHANNELS,
                rate = RATE,
                input = True,
                frames_per_buffer = chunk)


# setup the encoder ...
enc = encoder.StreamEncoder()
enc.set_channels(1)
#enc.set_bits_per_sample(wav.getsampwidth()*8)
enc.set_sample_rate(16000)
#enc.set_compression_level(0)

# initialize
if enc.init_stream(write) != encoder.FLAC__STREAM_ENCODER_OK:
    print "Error"
    sys.exit()

# start encoding !
stt.start()
nsamples = 512
while 1:
    data = stream.read(nsamples)
    if not data:
        enc.finish()
        break
    enc.process(data, nsamples)
    #sleep(.001)

stt.stop()

Could someone please tell me how to install the flac.encoder module (xp or linux). I have been searching all over for it. Thanks.

EDIT: okay got the script working after installing pyflac. Should have read the line on top and saved some time.
But this is what I get (I am using the simple API access key for the web)
403 Forbidden!!! Your client does not have permission to get URL /speech-api/full-duplex/v1/down?pair=.... from this server. That’s all we know.

Has anyone got it to work recently?

Is it a key you got from Google APIs after enabling Google Search ON? Because that is the key you require if I'm not wrong.

Also please help me out with installing Pyflac, i'm not sure on how to use SWIG and Pyflac

You need the latest pyflac, so it is recommended to build and install it from source. This is google developer key and as far as i remember you need to request access to google voice streaming api.

is this the pyflac you're using? https://github.com/dsully/pyflac (master, commit 578489 latest as of now) When I try to compile it on ubuntu 13.04, I get the errors here

https://gist.github.com/sid137/6546329

Here are my library versions:
ii swig 2.0.8-1ubuntu3 amd64 Generate scripting interfaces to C/C++ code
ii swig2.0 2.0.8-1ubuntu3 amd64 Generate scripting interfaces to C/C++ code
ii libflac-dev:amd64 1.2.1-6build1 amd64 Free Lossless Audio Codec - C development library

I was trying to adapt this to skip the recording part and transfer audio files that are already in the flac format. It seems like I shouldn't need pyflac or pyaudio, but all I've gotten are empty responses. I'm using a 40 minute *.mp3 file recorded at 44100 Hz that I've split into 15 second clips and converted to *.flac. I've tried sending it all at once and in 512-byte chunks. I've also tried using http://audio.online-convert.com/convert-to-flac to convert it to 8000 Hz, mono. Any ideas?