app.py

import time

import torch
import librosa

import gradio as gr

from transformers import AutoModelForCTC, Wav2Vec2BertProcessor

model_name = "Yehor/w2v-bert-2.0-uk"
device = "cpu"
max_duration = 30

asr_model = AutoModelForCTC.from_pretrained(model_name).to(device)
processor = Wav2Vec2BertProcessor.from_pretrained(model_name)

audio_samples = [
    "sample_1.wav",
    "sample_2.wav",
    "sample_3.wav",
    "sample_4.wav",
    "sample_5.wav",
    "sample_6.wav",
]

description_head = """
# Speech-to-Text for Ukrainian

## Overview

This space uses https://huggingface.co/Yehor/w2v-bert-2.0-uk model that solves 
a Speech-to-Text task for the Ukrainian language.
""".strip()

description_foot = """
## Community

- Join our Discord server - https://discord.gg/yVAjkBgmt4 - where we're talking about Data Science, 
Machine Learning, Deep Learning, and Artificial Intelligence.

- Join our Speech Recognition Group in Telegram: https://t.me/speech_recognition_uk
""".strip()


def inference(audio_path, progress=gr.Progress()):
    gr.Info("Starting process", duration=2)

    progress(0, desc="Starting")

    duration = librosa.get_duration(path=audio_path)
    if duration > max_duration:
        raise gr.Error("The duration of the file exceeds 10 seconds.")

    paths = [
        audio_path,
    ]

    results = []

    for path in progress.tqdm(paths, desc="Recognizing...", unit="file"):
        t0 = time.time()

        audio_duration = librosa.get_duration(path=path, sr=16_000)
        audio_input, _ = librosa.load(path, mono=True, sr=16_000)

        features = processor([audio_input], sampling_rate=16_000).input_features
        features = torch.tensor(features).to(device)

        with torch.inference_mode():
            logits = asr_model(features).logits

        predicted_ids = torch.argmax(logits, dim=-1)
        predictions = processor.batch_decode(predicted_ids)

        elapsed_time = round(time.time() - t0, 2)
        rtf = round(elapsed_time / audio_duration, 4)
        audio_duration = round(audio_duration, 2)

        results.append(
            {
                "path": path.split("/")[-1],
                "transcription": "\n".join(predictions),
                "audio_duration": audio_duration,
                "rtf": rtf,
            }
        )

    gr.Info("Finished...", duration=2)

    result_texts = []

    for result in results:
        result_texts.append(f'**{result["path"]}**')
        result_texts.append("\n\n")
        result_texts.append(f"> {result['transcription']}")
        result_texts.append("\n\n")
        result_texts.append(f'**Audio duration**: {result['audio_duration']}')
        result_texts.append("\n")
        result_texts.append(f'**Real-Time Factor**: {result['rtf']}')

    return "\n".join(result_texts)


demo = gr.Blocks(
    title="Speech-to-Text for Ukrainian",
    analytics_enabled=False,
)

with demo:
    gr.Markdown(description_head)

    gr.Markdown(f"## Demo (max. duration: **{max_duration}** seconds)")

    with gr.Row():
        audio_file = gr.Audio(label="Audio file", type="filepath")
        transcription = gr.Markdown(
            label="Transcription",
            value="Recognized text will appear here. Use **an example file** below the Recognize button,"
            "upload **your audio file**, or use **the microphone** to record something...",
        )

    gr.Button("Recognize").click(inference, inputs=audio_file, outputs=transcription)

    with gr.Row():
        gr.Examples(
            label="Choose an example audio", inputs=audio_file, examples=audio_samples
        )

    gr.Markdown(description_foot)

if __name__ == "__main__":
    demo.launch()