The -v option sets verbose mode and must be the first argument if present. In verbose mode, all spoken text is echoed to stdout. Since talk does some limited filtering of the input data, it can be useful to collect the interpreted text.
Any number of files can be read via the -f option. However you must specify a file name for each occurance of -f.
A single - means read from stdin (at that position in the file list. You must include the - to read stdin. No arguments will cause talk to exit.
Any remaining arguments on the command lines are concatenated into a single line (hence single sentence) and "spoken" after any files or stdin is read.
Talk tries to find speak.tos by first looking for it in the current directory, then via the environment variable SPEAK (which must contain the full path, name, and extension), and finally by searching the PATH for "speak.tos". If it fails, it quits.