[Grml] Customizing GRML to start speech as early as possible

First thing, I want to say, I am really fond of GRML because of it's accessibility features. As  the Manager of High Performance Computing at the Math Department of the University of Wisconsin, I was responsible for a beowulf cluster of 6 to 8 small super computers and about 100 Linux desktops. GRML saved my bacon many times. However, I do have a dream. I would like  to building a bootable ISO that starts speech automatically during boot.

To start speakup, you have to load a kernel module, speakup_soft. Then you have to run a program, espeakup. I know it is possible to configure that in an initrd file. But I have no idea how to go about doing it with GRML.

Any idea would be most greatly appreciated.