Spaces:
Running
Running
from transformers import pipeline | |
from datasets import load_dataset | |
import gradio as gr | |
import os | |
atco2 = load_dataset('jlvdoorn/atco2-asr', split='validation') | |
atcosim = load_dataset('jlvdoorn/atcosim', split='validation') | |
num_examples = 3 | |
examples_atco2 = [ [{'sampling_rate': atco2[i]['audio']['sampling_rate'], 'raw': atco2[i]['audio']['array']}, False, 'large-v3'] for i in range(num_examples)] | |
#examples_atcosim = [ [{'sampling_rate': atcosim[i]['audio']['sampling_rate'], 'raw': atcosim[i]['audio']['array']}, False, 'large-v3'] for i in range(num_examples)] | |
examples = examples_atco2 #+ examples_atcosim | |
whisper = pipeline(model='jlvdoorn/whisper-large-v3-atco2-asr-atcosim') | |
def transcribe(audio, model_version): | |
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(source='upload', interactive=True), | |
gr.Checkbox(label='Transcribe only', default=False), | |
gr.Dropdown(choices=['large-v2', 'large-v3'], value='large-v3', label='Whisper model version') | |
], | |
outputs = [gr.Textbox(label='Transcription'), gr.Textbox(label='Callsigns, commands and values')], | |
title = 'Whisper ATC - Large v3', | |
description = 'Transcribe ATC speech', | |
# examples = examples, | |
) | |
mic_iface = gr.Interface( | |
fn = transcribe, | |
inputs = [gr.Audio(source='microphone', type='filepath'), | |
gr.Checkbox(label='Transcribe only', default=False), | |
gr.Dropdown(choices=['large-v2', 'large-v3'], value='large-v3', label='Whisper model version') | |
], | |
outputs = [gr.Textbox(label='Transcription'), gr.Textbox(label='Callsigns, commands and values')], | |
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') | |