There is software available that will allow for some degree of "Centre Channel Extraction" or "Centre Channel Isolation" if the original recording is in stereo and, ideally, is in wav format, rather than mp3. Also all the vocals must be at, or very close to the absolute centre of the stereo image.
Essentially, what this software does is to "boost" the centre of the stereo image, whilst "cutting" the elements that are predominantly or entirely in the right or left channels. It can also do the reverse; "cut" the centre and "boost" the surround.
However, typical recordings are likely to have elements of the whole sound that "share" the centre of the sound image, e.g. the kick drum; no software can satisfactorily discriminate between vocals and non-vocals so those other "centre" elements will also be affected by the cutting or boosting of the stereo centre. Similarly, some elements of the vocal may well also be in the right or left channels, or both, most notably perhaps any reverb or effects.
If you Google Centre Channel Extraction (as I just did) almost every item on the first page mentions the only software I know that comes anywhere near to achieving what you want to do; it is an effect within a constituent part of the Adobe CC Suite, namely Audition. I have this software and can vouch for the fact it is the best for performing this type of work that I have encountered; but it is not, and arguably never can be, "perfect".
I was wondering if there was any way to isolate the vocals of an mp3 file so that they just play?
Roland had an incredible solution R-MIX but it was $400 several years ago. You can find it now and then on ebay. "Roland R-Mix Audio Processing Windows PC 10 8 7". You can pull out individual instruments as well.
I did a quick Google search on that Roland unit and found a series of reviews which were not wholly complimentary, but I see it is now available as a VST, apparently, see this page. And the price appears to be rather less than it was before! But this Amazon UK page lists it (apparently the full standalone version) at considerably more!
Plug-in specialist Zynaptiq is currently working on fine-tuning a solution for the "unmixing" of a song into its constituent stem tracks, which is going to be integrated with the next generation of ACID Pro. It has plenty of artificial intelligence under the hood, and is looking somewhat promising, given the technical confines within which something like this is actually possible. Who knows - maybe it's going to find its way into Music Maker at some point in time. 🙂