I have control over everything. I get the same results whether using studio produced voice tracks or those I created on my own. The artifacts are from the software not the recording process but I understand that I need a pristine source to get the best results. It is obvious to me that the source of this artifact is in the software manipulation stage not the creation of sound file stage.
I think that you should take a bit more care with your descriptions.
I am guessing that the "creation of sound file stage" refers to the input file (before transformation), not the output file (after transformation). And "the software" you say is the source of the artifacts refers just to the sound authoring package.
(I was also not always sure when you were referring to software as artifact removal software, and when you were referring to the software that was creating the artifacts. Sometimes, you just said, "the software".)
Algorithms for transforming sound exist on sound authoring software, on sound card interfaces and on sound card on-board processors (hardware). The authoring software can query a software interface driver, and if it has the capability, ask it to perform elements of the algorithm. Even though the requested function may be defined in some international standard (guaranteeing a well-defined outcome), there are various standards to choose from. So, the processing can depend on which standard the driver conforms with. Similarly, the driver can ask the sound card's on-board processor to do the task. So, the algorithm can end up being shared among different hardware and software components on your system, and be hardware and configuration dependent.
If your processing is well-defined by international standards, then you can be guaranteed system independence for basic transformations (like converting file format), or high fidelity for complex ones. This is because, even if the requirement is subjective or otherwise imprecise, hardware manufacturers and software developers are competing to one standard and subjectively trying to make their product sound good. So, you can take the standardized output file from one system and, although it may sound slightly different on another system, it will be as good as the producer of that system intended.
On the other hand, if you are dealing with transformations that have no standards, one manufacturer/software developer cannot guarantee that what sounds natural on his test system won't sound "mechanical" on another. I do not see any way that you can prevent your transformation software from being completely system dependent. There is no standard for stakeholders to test against to ensure that their product conforms with any other.
Researchers who pioneer new transformation technology will have little choice but to pick one system and make it work well on that system. If you change system, all you can do is tweak everything from interface software to hardware, to try to optimize the technology.
In summary, unless you know for sure how the software works, it may not be so obvious where sound artifacts are coming from.
Perhaps you can get a free trial of that $3000 software and try it out on your computer, and see if it is still as good as advertised.