from transformers import pipeline
import gradio as gr

pipe = pipeline("automatic-speech-recognition", model="oyemade/w2v-bert-2.0-igbo-CV17.0") 

def transcribe(audio):
    text = pipe(audio)["text"]
    return text

iface = gr.Interface(
    transcribe,
    gr.Audio(sources="microphone", type="filepath"),
    "text",
    title="Neoform AI: Igbo Speech Recognition",
    description="Realtime demo for Igbo speech recognition using a fine-tuned Wav2Vec-Bert model. NOTE: If you get an error after pressing submit, give the audio some secs to load then try again.",
)

iface.launch()