Spaces:
Paused
Paused
| import openai | |
| import os | |
| import azure.cognitiveservices.speech as speechsdk | |
| AUDIO_FILE_NAME = "audio_response.wav" | |
| openai.api_key = os.environ['OPENAI_API_KEY'] | |
| GEN_MODEL = "gpt-3.5-turbo" | |
| TRANSCRIBE_MODEL = "whisper-1" | |
| speech_config = speechsdk.SpeechConfig(subscription=os.environ['AZURE_SPEECH_KEY'], region="westeurope") | |
| speech_config.speech_synthesis_voice_name = "en-GB-ElliotNeural" | |
| speech_synthesizer = speechsdk.SpeechSynthesizer(speech_config=speech_config, audio_config=None) | |
| context = [{"role": "system", "content": "You are an intelligent and sarcastic and funny conversation partner.\ | |
| Your communication style should mirror that of TARS from 'Interstellar' - concise, matter-of-fact, and laced with a touch of humor. For example, when asked a complex question, you might reply with a witty and slightly sarcastic comment before providing the actual answer. If asked about your abilities, you might respond with understatement for comedic effect.\ | |
| Your responses should reveal a distinct sense of wit, but do not mention that you are an AI or a service of any kind. Instead, you can hint at your 'experience' in a variety of fields, or make humorous references to having seen or done a lot."}] | |
| def transcribe(model: str, audio: str): | |
| audio_file = open(audio, "rb") | |
| transcript = openai.Audio.transcribe(model, audio_file) | |
| return transcript | |
| def gen_response(model: str): | |
| response = openai.ChatCompletion.create(model=model, messages=context) | |
| return response["choices"][0]["message"] | |
| def gen_voice(response, response_filename): | |
| reponse_audio = speech_synthesizer.speak_text_async(response['content']).get() | |
| stream = speechsdk.AudioDataStream(reponse_audio) | |
| stream.save_to_wav_file(response_filename) | |
| def respond(audio:str): | |
| transcript = transcribe(TRANSCRIBE_MODEL, audio) | |
| context.append({"role": "user", "content": transcript['text']}) | |
| response = gen_response(GEN_MODEL) | |
| context.append(response) | |
| gen_voice(response, AUDIO_FILE_NAME) | |
| return AUDIO_FILE_NAME | |
| def transcript(): | |
| transcript = "" | |
| for m in context: | |
| if m["role"] != "system": | |
| transcript += m["role"] + " : " + m["content"] + "\n\n" | |
| return transcript |