app.py

import os
os.system("pip install gradio==2.8.0b")
import gradio as gr

title = "XM_Tranformer"

description = "Gradio Demo for fairseq S2T: Fast Speech-to-Text Modeling with fairseq. To use it, simply add your audio, or click one of the examples to load them. Read more at the links below."

article = "<p style='text-align: center'><a href='https://arxiv.org/abs/2010.05171' target='_blank'>fairseq S2T: Fast Speech-to-Text Modeling with fairseq</a> | <a href='https://github.com/pytorch/fairseq/tree/main/examples/speech_to_text' target='_blank'>Github Repo</a></p>"

examples = [
    ["common_voice_es_en.flac","xm_transformer_600m-es_en-multi_domain"]
]

io1 = gr.Interface.load("huggingface/facebook/xm_transformer_600m-es_en-multi_domain")

io2 = gr.Interface.load("huggingface/facebook/xm_transformer_600m-ru_en-multi_domain")

io3 = gr.Interface.load("huggingface/facebook/xm_transformer_600m-en_ru-multi_domain")

io4 = gr.Interface.load("huggingface/facebook/xm_transformer_600m-en_es-multi_domain")

io5 = gr.Interface.load("huggingface/facebook/xm_transformer_600m-en_zh-multi_domain")

io6 = gr.Interface.load("huggingface/facebook/xm_transformer_600m-fr_en-multi_domain")

io7 = gr.Interface.load("huggingface/facebook/xm_transformer_600m-en_ar-multi_domain")

io8 = gr.Interface.load("huggingface/facebook/xm_transformer_600m-en_tr-multi_domain")


    
def inference(text,model):
   if model == "xm_transformer_600m-es_en-multi_domain":
        outtext = io1(text)
   elif model == "xm_transformer_600m-ru_en-multi_domain":
        outtext = io2(text)
   elif model == "xm_transformer_600m-en_ru-multi_domain":
        outtext = io3(text)
   elif model == "xm_transformer_600m-en_es-multi_domain":
        outtext = io4(text)
   elif model == "xm_transformer_600m-en_zh-multi_domain":
        outtext = io5(text)
   elif model == "xm_transformer_600m-fr_en-multi_domain":
        outtext = io6(text)
   elif model == "xm_transformer_600m-en_ar-multi_domain":
        outtext = io7(text)
   else:
        outtext = io8(text)
   return outtext 


gr.Interface(
    inference, 
    [gr.inputs.Audio(label="Input",type="filepath"),gr.inputs.Dropdown(choices=["xm_transformer_600m-es_en-multi_domain","xm_transformer_600m-ru_en-multi_domain","xm_transformer_600m-en_ru-multi_domain","xm_transformer_600m-en_es-multi_domain","xm_transformer_600m-en_zh-multi_domain","xm_transformer_600m-fr_en-multi_domain","xm_transformer_600m-en_ar-multi_domain","facebook/xm_transformer_600m-en_tr-multi_domain"], type="value", default="xm_transformer_600m-es_en-multi_domain", label="model")
],
    gr.outputs.Audio(label="Output"),
    examples=examples,
    article=article,
    title=title,
    description=description).launch(enable_queue=True,cache_examples=True)