from transformers import pipeline
import gradio as gr

whisper = pipeline(model='jlvdoorn/whisper-large-v3-atco2-asr-atcosim')

def transcribe(audio):
    if audio is not None:
        return whisper(audio)['text']
    else:
        return 'There was no audio to transcribe...'

file_iface = gr.Interface(
    fn = transcribe,
    inputs = gr.Audio(sources='upload', interactive=True),

    outputs = gr.Textbox(label='Transcription'),
    title = 'Whisper ATC - Large v3',
    description = 'Transcribe ATC speech',
)

mic_iface = gr.Interface(
    fn = transcribe,
    inputs = gr.Audio(sources='microphone', type='filepath'),

    outputs = gr.Textbox(label='Transcription'),
    title = 'Whisper ATC - Large v3',
    description = 'Transcribe ATC speech',
)

demo = gr.TabbedInterface([file_iface, mic_iface], ["File", "Microphone"])
demo.launch(server_name='0.0.0.0')