Generating tiff/box file from original image and correct text lines of image

[enkhbaata...@unimedia.co.jp]

As i understood from wiki, training tesseract lstm requires tiff/box file. I have a dataset which includes small images of text line and texts those images recognized correctly. After researching few tips from tutorials i need to run tesseract with nobatch, makebox options to generate boxfile but it only uses my images to generate box file and can not recognize correct text with it. So i have to edit box file manually using jtessboxeditor even i have already correct text of those images. Is there any possible way to generate box file using image and text file without tesseract command?