Native Multitrack Exports Instead of AI Stem Separation

I have a question about how Songer generates and exports instrument stems.

Since Songer generates the music, I had assumed that the individual instrumental parts were available internally and that the Stems Studio feature simply exported those original tracks.

However, after examining several exported stems, I've noticed some apparent overlap between instruments, particularly guitar and keyboard. This makes me wonder whether Stems Studio is actually using AI source separation on the completed stereo mix rather than exporting the original generated tracks.

Could someone from the Songer team clarify how the process works?

If stems are currently created through source separation, would it be possible to consider native multitrack exports in the future? Having access to the original generated instrument tracks would provide much cleaner separation and considerably more flexibility for musicians who want to work with their compositions in a DAW.

I appreciate what Songer has made possible for independent music creators, and this would be a very valuable addition to an already impressive platform.