Spaces:
Build error
Build error
| import os | |
| import gradio as gr | |
| import whisper | |
| import requests | |
| import tempfile | |
| from neon_tts_plugin_coqui import CoquiTTS | |
| # Whisper: Speech-to-text | |
| model = whisper.load_model("base") | |
| # LLM : Bloom as inference | |
| API_URL = "https://api-inference.huggingface.co/models/bigscience/bloom" | |
| HF_TOKEN = os.environ["HF_TOKEN"] | |
| headers = {"Authorization": f"Bearer {HF_TOKEN}"} | |
| #Language covered in Bloom : en, fr, esp, arb, hn, portu, Indonesian, Vietnamese, Chinese, tamil, telugu, bengali | |
| # Text-to-Speech | |
| LANGUAGES = list(CoquiTTS.langs.keys()) | |
| print(f"Languages for Coqui are: {LANGUAGES}") | |
| #Languages for Coqui are: ['en', 'es', 'fr', 'de', 'pl', 'uk', 'ro', 'hu', 'el', 'bg', 'nl', 'fi', 'sl', 'lv', 'ga'] | |
| coquiTTS = CoquiTTS() | |
| # Driver function | |
| def driver_fun(audio) : | |
| text1, lang = whisper_stt(audio) | |
| #text1 = model.transcribe(audio)["text"] | |
| text2 = lang_model_response(text1, lang) | |
| speech = tts(text2, lang) #'en') | |
| return text1, text2, speech | |
| # Whisper - speeech-to-text | |
| def whisper_stt(audio): | |
| print("Inside Whisper TTS") | |
| # load audio and pad/trim it to fit 30 seconds | |
| audio = whisper.load_audio(audio) | |
| audio = whisper.pad_or_trim(audio) | |
| # make log-Mel spectrogram and move to the same device as the model | |
| mel = whisper.log_mel_spectrogram(audio).to(model.device) | |
| # detect the spoken language | |
| _, probs = model.detect_language(mel) | |
| lang = max(probs, key=probs.get) | |
| print(f"Detected language: {max(probs, key=probs.get)}") | |
| # decode the audio | |
| options = whisper.DecodingOptions(fp16 = False, language=lang) | |
| result = whisper.decode(model, mel, options) | |
| # print the recognized text | |
| print(f"transcript is : {result.text}") | |
| return result.text, lang | |
| # LLM - Bloom Response | |
| def lang_model_response(prompt, language): | |
| print(f"*****Inside lang_model_response - Prompt is :{prompt}") | |
| p = """Question: How are you doing today? | |
| Answer: I am doing good, thanks. | |
| Question: """ | |
| if len(prompt) == 0: | |
| prompt = """Question: Can you help me please? | |
| Answer: Sure, I am here for you. | |
| Question: """ | |
| prompt = p + prompt + "\n" + "Answer: " | |
| json_ = {"inputs": prompt, | |
| "parameters": | |
| { | |
| "top_p": 0.90, #0.90 default | |
| "max_new_tokens": 64, | |
| "temperature": 1.1, #1.1 default | |
| "return_full_text": False, | |
| "do_sample": True, | |
| }, | |
| "options": | |
| {"use_cache": True, | |
| "wait_for_model": True, | |
| },} | |
| response = requests.post(API_URL, headers=headers, json=json_) | |
| #print(f"Response is : {response}") | |
| output = response.json() | |
| output_tmp = output[0]['generated_text'] | |
| print(f"Bloom API Response is : {output_tmp}") | |
| solution = output_tmp.split("Answer: ")[2].split("\n")[0] | |
| print(f"Final Bloom Response after splits is: {solution}") | |
| return solution | |
| # Coqui - Text-to-Speech | |
| def tts(text, language): | |
| print(f"Inside tts - language is : {language}") | |
| coqui_langs = ['en' ,'es' ,'fr' ,'de' ,'pl' ,'uk' ,'ro' ,'hu' ,'bg' ,'nl' ,'fi' ,'sl' ,'lv' ,'ga'] | |
| if language not in coqui_langs: | |
| language = 'en' | |
| with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as fp: | |
| coquiTTS.get_tts(text, fp, speaker = {"language" : language}) | |
| return fp.name | |
| gr.Interface( | |
| title = 'Testing Whisper', | |
| fn=driver_fun, | |
| inputs=[ | |
| gr.Audio(source="microphone", type="filepath"), #streaming = True, | |
| # "state" | |
| ], | |
| outputs=[ | |
| "textbox", "textbox", "audio", | |
| ], | |
| live=True).launch() | |