I am developing an app and I want to detect the language automatically and then print it.
My code:
with sr.Microphone() as source:
audio = r.listen(source)
try:
# Auto detect the language
print("You said: " + r.recognize_google(audio))
except sr.UnknownValueError:
print("Google Speech Recognition could not understand audio")
except sr.RequestError as e:
print("Could not request results from Google Speech Recognition service")
Hope you understood.
This may be too late for your answer. However, someone else may be looking for this in the future.
I have not found a way to auto detect the language using speech_recognition. The Google API does support multiple languages in the form of an alternate languages array in which it will attempt the different languages you've specified in order to provide a translation.
The way I overcame the issue of detecting a language is that I only have two languages. I could have more I only have two that I need. I use a wake word and command. So I can say (for example) "ok computer translate from Spanish" and the command parser determines that the intent is to translate from Spanish. I used snips for this but you could just do string tokenization. In any case, at that point, because I know the intent is "from Spanish" I explicitly set the language code to "es" like this:
said = r.recognize_google(audio, language="es")
My SpeechRecord class
import speech_recognition as sr
class SpeechRec:
#https://techwithtim.net/tutorials/voice-assistant/wake-keyword/
def record(self, lang='en'):
r = sr.Recognizer()
with sr.Microphone() as source:
audio = r.listen(source)
said = ""
try:
#can I detect the language?
if (lang == 'en') :
said = r.recognize_google(audio, language='en-US')
elif (lang == 'es') :
said = r.recognize_google(audio, language="es")
print(said)
except Exception as e:
if (str(e) != ""):
print("Exception: " + str(e))
return said.lower()
The listen loop - I call this from a Flask event but it'll work the same in a standalone application
WAKE = "computer"
while True:
text = SpeechRec().record()
language = "en"
if text.count(WAKE) > 0:
text = SpeechRec().record()
#here I have a call to determine the intent via Snips - I've removed that and just
#placed a text comparison for simplicity. Also, to note, using Snips I can reduce
#the wake word and command to remove two of these
#"text = SpeechRec().record()" lines
if (text == 'translate from spanish'):
text = SpeechRec().record('es')
else:
text = SpeechRec().record()
#at this point I do the translation and print the value from the translation
This is probably not the most elegant solution. I will probably rewrite this many times in the future. However, it works well for what I need at the moment.
I hope this helps answer your question.
