Is there a way to reproduce vision split more accurately?

[little-bird-vodka]

Thanks for releasing the code for composing vision split. We found that the background, font size and font style are chosen randomly in the code. Is there a way to reproduce the vision split more accurately? We wish each of our images could have the same layout as the corresponding image in the vision split.

Some images of vision split seem to be obtained through manual screenshots or taking photos of the monitor, disturbances they brought cannot be replicated. However, it would still be very helpful if we could at least align the font styles and background templates.