Right. So size only.
I am busy processing census images at the moment.
What I do is for each individual I enter Residence Date Place and if appropriate address in the description field. (FTM I found doesn't like you you having more than about 20,000 places.) I also enter occupation if found. The source for each is US Federal Census xxxx. The image comes nowhere near the program. If I want to look at the image, I go to my spreadsheet, find the number allocated such as 1940-1234.
What I would do in your position is to export a gedcom having deselected the include media box and load it into FTM as a new file with a new file name. I would then spend a couple of hours adding data to it and looking at what had happened to my sources and assessing how the program performed. If all went well I would adopt this as my main file and decide how I was going to deal with anything connected to an image in future.
I realise this even if it solves your problem is going to involve a lot of work, but equally if you jump ship to another program you are going to have a lot of work as well.