I'm not entirely certain what you mean by doing load testing using existing scripts so let me run this past you.
If by this you mean that you have a script project that does a functional/regression test of a web application and that you want to use that script project as the "execution" of the load against the store, the answer is "not really."
The LoadTesting feature of TestComplete is a traffic transponder/recorder. It records HTTP Requests and Responses and then plays them back as a virtual user. These requests and responses are ONLY network traffic. None of the mouse clicks, keyboard keystrokes, etc, are recorded during task recording. To that end, this means playback of load testing does not involve the GUI at all.
Regular scripts actually interact with the GUI and the application, executing mouse clicks and keystrokes. To have multiple virtual users on the same box doing that would not be possible.
So... there are two ways you can potentially accommodate this:
1) Using the Network Suites and distributed testing, execute your existing functional/regression scripts on multiple workstations or Virtual machines using TestComplete or TestExecute instances on those machines. This is expensive when it comes to environment because if you want to run a load of 100 users, you literally need 100 machines and virtual machines not to mention 100 licenses of TC or TE.
2) I THINK (although I have not done so) there is a way you can set up one machine to execute the scripts and use another machine to record the HTTP traffic. I believe it has to do with setting the proxy on the recording machine in some way. Perhaps someone else can chime in.
However, in all honesty, it's probably easiest to just manually record your HTTP tasks. There's two reasons for this:
1) Most of the time, functional scripts have a LOT of excess activity in them with clicking around the page, checking different options, entering multiple data records, as well as handling several different test cases. For a load test, you typically want to simulate a "real" user... what the typical activity would be on the site. So, it would be easiest to just record a simple user process as an HTTP task and assign that one process to multiple virtual users
2) Even after you record your traffic using your scripts, you still need to design your test (assigning tasks to virtual users, building your populations, etc). So, while you can use your scripts to generate the recorded traffic, it's not actually load testing...you're basically just automating one part of the recording that you need to do anyways for generating your load test.