app.py

# -*- coding: utf-8 -*-
import torch
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
import librosa
import numpy as np
from datetime import timedelta
import gradio as gr
import os

def format_time(seconds):
    td = timedelta(seconds=seconds)
    hours, remainder = divmod(td.seconds, 3600)
    minutes, seconds = divmod(remainder, 60)
    milliseconds = td.microseconds // 1000
    return f"{hours:02d}:{minutes:02d}:{seconds:02d},{milliseconds:03d}"

def estimate_word_timings(transcription, total_duration):
    words = transcription.split()
    total_chars = sum(len(word) for word in words)
    char_duration = total_duration / total_chars

    word_timings = []
    current_time = 0

    for word in words:
        word_duration = len(word) * char_duration
        start_time = current_time
        end_time = current_time + word_duration
        word_timings.append((word, start_time, end_time))
        current_time = end_time

    return word_timings

model_name = "Akashpb13/xlsr_kurmanji_kurdish"
model = Wav2Vec2ForCTC.from_pretrained(model_name)
processor = Wav2Vec2Processor.from_pretrained(model_name)

def transcribe_audio(file):
    speech, rate = librosa.load(file, sr=16000)
    input_values = processor(speech, return_tensors="pt", sampling_rate=rate).input_values

    with torch.no_grad():
        logits = model(input_values).logits

    predicted_ids = torch.argmax(logits, dim=-1)
    transcription = processor.batch_decode(predicted_ids)[0]
    total_duration = len(speech) / rate
    word_timings = estimate_word_timings(transcription, total_duration)

    srt_content = ""
    for i, (word, start_time, end_time) in enumerate(word_timings, start=1):
        start_time_str = format_time(start_time)
        end_time_str = format_time(end_time)
        srt_content += f"{i}\n{start_time_str} --> {end_time_str}\n{word}\n\n"

    output_filename = "output_word_by_word.srt"
    with open(output_filename, "w", encoding="utf-8") as f:
        f.write(srt_content)

    return transcription, output_filename

interface = gr.Interface(
    fn=transcribe_audio,
    inputs=gr.Audio(type="filepath"),
    outputs=[gr.Textbox(label="Transcription"), gr.File(label="Download SRT File")],
    title="Deng --- Nivîsandin ::: Kurdî-Kurmancî",
    description="Dengê xwe ji me re rêke û li Submit bixe ... û bila bêhna te fireh be .",
    article="By Derax Elî"
)

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