from transformers import pipeline
import gradio as gr
from pyctcdecode import BeamSearchDecoderCTC
from Aniemore import EmotionFromVoice

# emo = pipeline("audio-classification", model="Aniemore/wav2vec2-xlsr-53-russian-emotion-recognition", trust_remote_code=True)
emo = EmotionFromVoice()


def transcribe(audio):
    emotion = emo(audio)
    return emotion

def get_asr_interface():
    return gr.Interface(
        fn=transcribe, 
        inputs=[
            gr.inputs.Audio(source="microphone", type="filepath")
        ],
        outputs=[
            "textbox",
        ])
        
interfaces = [
    get_asr_interface()
]

names = [
    "ASR"
]

gr.TabbedInterface(interfaces, names).launch(server_name = "0.0.0.0", enable_queue=False)