app.py

import gradio as gr
import librosa
from asr import transcribe
from tts import synthesize

def identify(microphone, file_upload):
    LID_SAMPLING_RATE = 16_000

    if (microphone is not None) and (file_upload is not None):
        return "WARNING: Using microphone input. Uploaded file will be ignored."

    if (microphone is None) and (file_upload is None):
        return "ERROR: Provide an audio file or use the microphone."

    audio_fp = microphone if microphone is not None else file_upload
    inputs = librosa.load(audio_fp, sr=LID_SAMPLING_RATE, mono=True)[0]

    return {"Faroese": 1.0}

demo = gr.Blocks()

mms_transcribe = gr.Interface(
    fn=transcribe,
    inputs=[
        gr.Audio(source="microphone", type="filepath"),
        gr.Audio(source="upload", type="filepath"),
    ],
    outputs="text",
    title="Speech-to-text",
    description="Transcribe audio!",
    allow_flagging="never",
)

mms_synthesize = gr.Interface(
    fn=synthesize,
    inputs=[
        gr.Text(label="Input text"),
        gr.Slider(minimum=0.1, maximum=4.0, value=1.0, step=0.1, label="Speed"),
    ],
    outputs=gr.Audio(label="Generated Audio", type="numpy"),
    title="Text-to-speech",
    description="Generate audio!",
    allow_flagging="never",
)

mms_identify = gr.Interface(
    fn=identify,
    inputs=[
        gr.Audio(source="microphone", type="filepath"),
        gr.Audio(source="upload", type="filepath"),
    ],
    outputs=gr.Label(num_top_classes=1),
    title="Language Identification",
    description="Identify the language of audio!",
    allow_flagging="never",
)

with demo:
    gr.TabbedInterface(
        [mms_synthesize, mms_transcribe, mms_identify],
        ["Text-to-speech", "Speech-to-text", "Language Identification"],
    )

demo.launch()