from transformers import pipeline

import gradio as gr

#jonatasgrosman/wav2vec2-large-xlsr-53-spanish
asr = pipeline("automatic-speech-recognition", "jonatasgrosman/wav2vec2-large-xlsr-53-spanish")
classifier = pipeline("text-classification", "finiteautomata/beto-sentiment-analysis")


def speech_to_text(speech):
    text = asr(speech)["text"]
    return text


def text_to_sentiment(text):
    return classifier(text)[0]["label"]


demo = gr.Blocks()

with demo:
    audio_file = gr.Audio(type="filepath")
    text = gr.Textbox()
    label = gr.Label()

    b1 = gr.Button("Recognize Speech")
    b2 = gr.Button("Classify Sentiment")

    b1.click(speech_to_text, inputs=audio_file, outputs=text)
    b2.click(text_to_sentiment, inputs=text, outputs=label)

demo.launch(inline=false)