Could you fix the Stem analyze to make it faster because it truly is not on the fly fast enough for mixing anything
They cannot magically put in another SoC, the speed is what it is.
We don’t know what the limits of the processor are so that’s the question
I doubt that, in its current state, the SoC isn’t being used at 100% of its capacity during the separation process.
Perhaps they can optimize the code to slightly improve the process speed, but it will remain marginal anyway. At best, you might get a 10-15% improvement, maybe 20%… But don’t expect them to magically double the separation speed.
And even if doubling it were possible, it would still be unusable for on-the-fly use.
The RS1 wasn’t designed for on-the-fly stem workflows, but for a “pre-render in the background on the device and play it later” workflow.
Any improvement to make it faster will help
Can’t be certain but I would like to think the Engine team are working on speeding it up, it is possible any new devices that come out maybe able to render in real time
Still not possible with an RK3588; it lacks the necessary performance. At least, not with decent audio quality—free from residue or artifacts from the other stems.
We’ll have to wait for the next hardware platform (maybe).
Some of us have worked with this SoC since its release in 2021. 