| import torch | |
| import scipy | |
| import gradio as gr | |
| from transformers import set_seed, pipeline | |
| from transformers import VitsTokenizer, VitsModel | |
| from transformers import pipeline, AutoTokenizer, AutoModelForSeq2SeqLM | |
| from datasets import load_dataset, Audio | |
| import speech_to_text, text_to_speech, translation | |
| language_list = ['mos', 'fra', 'eng'] | |
| demo = gr.Blocks() | |
| mms_stt = gr.Interface( | |
| fn=speech_to_text.transcribe, | |
| inputs=[ | |
| gr.Audio(sources=["microphone", "upload"], type="filepath"), | |
| gr.Dropdown(language_list, label="Language") | |
| ], | |
| outputs="text", | |
| title="Speech-to-text" | |
| ) | |
| mms_tts = gr.Interface( | |
| fn=text_to_speech.synthesize_facebook, | |
| inputs=[ | |
| gr.Text(label="Input text"), | |
| gr.Dropdown(language_list, label="Language") | |
| ], | |
| outputs=[ | |
| gr.Audio(label="Generated Audio", type="numpy") | |
| ], | |
| title="Text-to-speech" | |
| ) | |
| mms_translate = gr.Interface( | |
| fn=translation.translation, | |
| inputs=[ | |
| gr.Textbox(label="Text", placeholder="Yaa sõama"), | |
| gr.Dropdown(label="Source Language", choices=["eng_Latn", "fra_Latn", "mos_Latn"]), | |
| gr.Dropdown(label="Target Language", choices=["eng_Latn", "fra_Latn", "mos_Latn"]) | |
| ], | |
| outputs=["text"], | |
| examples=[["Building a translation demo with Gradio is so easy!", "eng_Latn", "mos_Latn"]], | |
| title="Translation Demo", | |
| ) | |
| with demo: | |
| gr.TabbedInterface( | |
| [mms_translate, mms_tts, mms_stt], | |
| ["Translation", "Text-to-speech", "Speech-to-text"], | |
| ) | |
| demo.launch() |