Low latency+high quality harmonizer in Melda - IT's POSSIBLE!

[mccy]

For some time I had the idea for a lower latency harmonizer (for live-usage) in MXXX in my mind. What do you think about thisone? It needs Midiinput for the harmonies and blends between two harmonizers (one with lowest latency, the other with highest quality) via transient triggered Envelopes. And most important part: Do not report latency to DAW, so that the dry signal is not delyed anyway. Good idea?

EDIT: High CPU consumption solved! Thank you for your input!

Code: Select all

$eNrtWk2MI0cVLu-ugb9DcuEYWooidqUZ2e2d2UmUEKdmxrNrGO943c5MJxdU4y7bnWlXOf0zM0ZiIRKICxHcUJD4O5IbEoec+DnBBSQQOeSQhCghQoHDIiFxA96r6j+3vbPjXUMCch92u15Vv3rvez-1XnlIc9O2m7ZtBzwMXdEPCLW4x7shd1pe1HcFCaora4Q+SswS3ed+4EpB1gh5mNC644bSt9wv8ToLxiR4Yq2ycu3a4-kJEpjrFXPlifVrhG5JEfrS84BHXbBDjzvEBGrk+1yELZ+DAI1tEjTaxon0j25XK9W12zeYP5QCGPmGJ08Mj4VcdMdVQps8BD5bwI6NAu5YMOISRdxtWyPPDWGM+5vEIfjc+fYzhZlqPEPu1GiiKMJA7L0oHEUhiJZJWY6J1VnEq7OI65NEvUNA4hdCbWJeoM+hwDfZEGDKFG1uEroTed403QCZmiw4aghiXlJvsBe+6i0bwuGnFeSbG5vE-FcsSbOx3WiIEX5SsjeZOKqkglmp7bd8GQTyGLaTDmxfXXM2y7Ib5iY64xFMUME82Sd02x8f8JD2euAxgeJ+nYF+aAhXHKGBZzgOStLyZd9nw60BE32uBcTVE+QtMHdf+qg9Se2xq32gzUfSR6HVTMETKyuVSS9UBK1rR0rvkPmp6yBAs53mxtfOcBo7BQShGIOI4BzWUMpwIHgQEOe7r+Dz4xo9cIUjTzRsN5gA49NTF11X+iBY-BB63XedItHOGNbFMah0c7-Vrlv1DlgZbOfznuuBaIiC3ZKuADudEuerb332D5d+9-ca3fFYP9gk5svl3Owj7398na49-ExutmzTViXnsCXa4uwowBdrIE+UVPvMiziSYPUhE04ltuyj2lFygXp54PYHxosRA9zGV2BZJxI8AY4QqpeOMzXv5yE23bbaxO74fAjiWQMG8NrPSzkEpwasuCeBULYhMYwIbYR8WMmJgGOzMK4WxlcL47XCeL0wvlYYb2TjMjx2M-JCt8XAuVXqIgUC4KkjHiKqfMJDdAgZjUhw3ZOHzIPQ1EDflP6QeYQqe2jS5tjouX4QGqOEGXjiiIMlM4nU8lzKu-5sQ8cHsqhmbIFuAY+2PMG4gtGu7B7p4ASvSPi3MTpzWdxuZTunr5k6B6jOhARE5zjFBzJY+QwGKqkYfcwqkPNOJ9i8V6OahfPXW6898vW--LI2wZYqt8jc2tZuQi3BRh2JAWcrH0+HmJCTUJkZPHrbP9boOIdlsvKb0yvv5INw2gvMxOg6Yy-Y5nu+i2dNJvR5jF5ahNG1OrHRzjR8EZGNBJE4zd8NEjy-giIikJeODIGfT2GxzXsMNopleedzz8Lzeu2eYaFLIcjJhcDA7VHOCgk6LuSylGCC5LAkHedQTkhXIVtChiT3hSwwcQzTWDUsGcFbUrQBQaFAiwpmEfMQQ9Kf0ohJl0xETP00bKs0v39oACSr3-j0G7-9x7u1V--2++bhQ3+uPf7TXx--7Acf1BLApj06zWM7qLYIjWDg9jLPzs6JD8O51ajFBPf0cRyXpFhYPZAxtHMaIzfsDmBYUH2uADDTnPAFzkdGT7MKFg-gZETMdTCUFopWQc-50KomaB24Tjj4KLlZWstt7Gxs7+ws1sNibSei-c3paH+ndg-4ribwwcnoj1ThOheELe53oXsLjMurZqXymBFKA-+-ssjDKAZyYwfrvgcGcipv5jSfz-XWEuxoFMrVEIvc-0qQEnJOJ1wsTqqKvwdE0Eh0Q-eY625et2BoxW1+7HZ5oWfHibj7z39V7HufGj39VPD0ZZ+-GLmwwMDm0XCxKLwC9b3mWSW2iogmdEwyApfERjLXHJmqBC-nu++Lc3ffkGpULZr02SRrwT9WaMEvLlvwZQv+YbbgQ1ckl2XLDnzZgS878GUH-r-Qgc8obZYd+LIDX3bgyw582YEvO-BlB-7-1IGr378vxRZss9CV+dZbEfK-eH8qfoN+AoCG4UQzfkEP1WQF9cpmTezHs1kTZ8--4-ld23Tdemm5M2vcX189CUwJgflEDEzTPUUnyYBRBARGqbQFThImGmV-GJDCdOkeMKVAXCgAUTpb8UXfH0wiQBCBT56NQDVzjYuTOl8s3tOc5RqlomvMZf-FwzBHm9p+7hznwnRG+8+lrtgwkKp0y4hZbF960ZDPm9kTHVUb-pHWsXpuHcv27s4emBP+rSZKqVSaXpZM-zFUfPoQ2vGZCFyYg+8iD5OOyJWM+VngWhfgdF3uGJdZGLLukSGFh-dAd9U3d++UXd3Y1DlWbAilik1OszZ0ZSzI69puWrtc9KEmcj7zwepLr3znJyhWxG9IT50z8Q6W63AICX0-mQ6garpFnN88+ZXe93-4vmppAJcMlmT7b7HPr9g-+nJtxva4zS6HDyaO7i4bE0e+8dL6F1-9Xi1mo28gMuzUqiIx3iAmP+88+fbey7+qUSsKQpbVUATtom6-4BiAdXgBcuu1f-6c4mkqxZ6Aqi5kPrQhEN0eKqLpvR4JGn0hfV64PqNWp95KrheTyzQ80yPHlRkeHajk+phJSrQzAF8ZIMaBWkQyCmzvHCiRWr-IUWHzTPzzGFZTOu4wjromExHGkrLs2Q4AOfAF6N5Rl6RMAG0sty+QwUxlbWssuoTGrNQd611jMa4RIALj8uGsANSTm6Au1If5i8HZFZMdp+wx3iQ2RE-CEZCQwKhgbrsV+SMZ4B1nQwShDykAjuKyrRpbWAxVLeui2GVb3UChcnGAJ+mNPDhLvDiORcd3fwSWQFbmBbvJQt89zSbnqNRu-xvv1iku

[mccy]

Adding a mixer in MXXX adds 0.2% CPU usage on my MAC Studio Ultimate. Is this real?

[mccy]

Sometimes I feel like an alien... Is this irrelevant?

Same principle may be even more easily applied to Mautopitch, one would get a latencyfree Autotuner... but who cares... just me.

[MeldaProduction]

Well, I don't really understand these settings . You are switching between to harmonizers using a modulator, but why? Is it the initial stage? It indeed seems a bit more coherent, but I generally don't mind if the backing vocals are a little late.

What I don't understand at all is the feedback there. I takes a lot of CPU and it seems that it is somehow preprocessing the signal for the modulator? It seems to me it works the same without it here. But anyways just increase the delay so that it doesn't kill the CPU.

[mccy]

Well, I don't really understand these settings . You are switching between to harmonizers using a modulator, but why? Is it the initial stage? It indeed seems a bit more coherent, but I generally don't mind if the backing vocals are a little late.

What I don't understand at all is the feedback there. I takes a lot of CPU and it seems that it is somehow preprocessing the signal for the modulator? It seems to me it works the same without it here. But anyways just increase the delay so that it doesn't kill the CPU.

You're right. One can live without those low latency harmony voices most of the time, but the high quality harmonizer settings are too late for my taste.
With this method only the first few milliseconds go through the faster but less quality harmonizer. For voice it doen't matter soundwise for the first milliseconds, but for long notes now the quality is much higher, using the high quality higher latency harmonizer for long notes...

It becomes more interesting building a realtime Autotuner this way... With autotune you allways delay the whole signal which is no so good for many live situations.

Is there another method to let just the transients control the ratio? I used the feedback to use only the transient for firing up the Envelope because the modulator lets me choose feedback as input... with the sidechain input I have not the possibility to use only the Transient signal, right?

Ah, should I set the transient modulation to 100% in the detector?

EDIT: THANK YOU, NOW IT WORKS!

Another method would be to fade in the first Transient, so that delayed Transients don't disturb.

[MeldaProduction]

Aaaha, got it. The problem is that the transient is kinda the most important thing there, so I'm not sure if that's the solution, but it's an interesting idea indeed.