import gradio as gr
import librosa
from asr import transcribe, ASR_EXAMPLES, ASR_LANGUAGES, ASR_NOTE
from tts import synthesize, TTS_EXAMPLES, TTS_LANGUAGES
from lid import identify, LID_EXAMPLES
mms_transcribe = gr.Interface(
fn=transcribe,
inputs=[
gr.Audio(),
gr.Dropdown(
[f"{k} ({v})" for k, v in ASR_LANGUAGES.items()],
label="Language",
value="eng English",
),
],
outputs="text",
examples=ASR_EXAMPLES,
title="Speech-to-text",
description=(
"Transcribe audio from a microphone or input file in your desired language."
),
article=ASR_NOTE,
allow_flagging="never",
)
def tts_wrapper(text, lang, speed):
audio, filtered_text, s3_url = synthesize(text, lang, speed)
return audio, filtered_text, s3_url
mms_synthesize = gr.Interface(
fn=tts_wrapper,
inputs=[
gr.Text(label="Input text"),
gr.Dropdown(
[f"{k} ({v})" for k, v in TTS_LANGUAGES.items()],
label="Language",
value="eng English",
),
gr.Slider(minimum=0.1, maximum=4.0, value=1.0, step=0.1, label="Speed"),
],
outputs=[
gr.Audio(label="Generated Audio"),
gr.Text(label="Filtered text after removing OOVs"),
gr.Text(label="S3 URL or Status"),
],
examples=TTS_EXAMPLES,
title="Text-to-speech",
description=("Generate audio in your desired language from input text."),
allow_flagging="never",
)
mms_identify = gr.Interface(
fn=identify,
inputs=[
gr.Audio(),
],
outputs=gr.Label(num_top_classes=10),
examples=LID_EXAMPLES,
title="Language Identification",
description=("Identity the language of input audio."),
allow_flagging="never",
)
tabbed_interface = gr.TabbedInterface(
[mms_transcribe, mms_synthesize, mms_identify],
["Speech-to-text", "Text-to-speech", "Language Identification"],
)
with gr.Blocks() as demo:
gr.Markdown(
"MMS: Scaling Speech Technology to 1000+ languages demo. See our blog post and paper.
"
)
gr.HTML(
"""Click on the appropriate tab to explore Speech-to-text (ASR), Text-to-speech (TTS) and Language identification (LID) demos."""
)
gr.HTML(
"""You can also finetune MMS models on your data using the recipes provides here - ASR TTS"""
)
gr.HTML(
"""
for more control and no queue."""
)
tabbed_interface.render()
gr.HTML(
"""
"""
)
if __name__ == "__main__":
demo.queue()
demo.launch(show_error=True)