"Master" should be all PARTs in the Performance (a mix of PARTs 1-16) and perhaps AD (audio) as well.

A/D input should be a modulation source. If you set the audio output (from Cubase, aka Montage audio input) to target Digital L/R - this can act as a modulation source. If you pick Assign L/R or Main L/R - these cannot be used as modulation sources as they are routed directly to outputs without funneling through what is considered the "AD Input".

You can setup the output routing of Digital In to be "OFF" so you do not hear the recording on output. This is most likely your desired configuration.

I am working on a song that requires the use of a vocoder.

In Cubase, I have recorded an audio track of me speaking into the microphone. I also have a MIDI track with the notes I want in the song. I want to set things up so that the audio track is used as the Modulator input to the Montage Vocoder effect, and the MIDI track controls the pitch of the sound.

Is it possible to set things up like this? If so, how?

Thanks for the question. Yes, it is... here’s how and here’s what to look out for...

What’s in play here is you want to work the Vocoder feature from the *opposite direction*... the Vocoder is designed to take real-time input from the microphone or from a MONTAGE Part’s Output as the Source and it applies the pitch input from the keys being pressed... Running the sound through the band pass filters that are extracting the vocal formants.

You want to playback the source from a pre-recorded track and have the note-on data (that will supply the pitch input) also playback from another pre-recorded track.

You can understand the design concept is for both input and control to be real-time and simultaneous in order to successfully ‘marry’ each syllable to a key-on event. This marriage being essential to intelligibility, you’ll want to ensure the coordination/synchronicity between these two, (routing issues aside)... the routing is not the difficult part of this equation. The difficulty will be in not being able to interact with the encoding process. This may or may not be an issue for your end goal — but it is something to be aware of in this scenario.

First Hurdle: Getting the Audio to the Vocoder
There are a couple of methods to get the Cubase Audio Track into the Vocoder. To properly trigger the Vocoder, you’ll need to get the audio into the A/D Input or enter the audio via a MONTAGE PART Output (as a Waveform).

Either real-time input from the A/D In or real-time input from a MONTAGE Part’s Output can be used to trigger the Vocoder.
__ Routing from Cubase to the A/D Input would involve D-to-A and then the MONTAGE A-to-D conversion, and is just untenable from a practical standpoint.
__ Export the Audio Track as “stem”, (or as a series of audio clips) @ 24-bit/44.1kHz. Create an AWM2 Part referencing this data as its ELEMENT’s Waveform. The Output of that ‘audio’ Part can be Source Input to the Vocoder.

Second Hurdle: Triggering the Note information
Routing MIDI Tracks data to the MONTAGE. One MIDI track with a Note-On event to playback the Vocal audio Part. The other track to enter the pitch data for the resulting encoded sound. The MIDI Track can be used to “play” the Vocoded Part.
__ Go to the “Effect” > “Routing” screen of the Vocoder Part and set data you want to be the pitch control as the Side Chain Input to the Vocoder. For example, the data you wish to use as Control is Track 12, Set Part 12 as the Input to the Vocoder.

The Vocoder is assigned as an Insertion Effect on the Synth Sound.
The Vocal Audio Part is the Side Chain Input to the Vocoder.

Wildcard: Things that could be an issue
Do the MIDI Note events synchronize with the Audio? As long as you can guarantee this, you can effectively do exactly what you propose. In normal use, the player sings or speaks into the microphone at the same time they play a note on the keyboard. If these are not coordinated/synchronized properly, the intelligibility will suffer. (Some syllables may get completely lost).

No sound is produced unless the source input occurs while a Note-On is occurring. It takes a bit of practice doing it normally, but rather quickly you figure out what you can and can’t do, as you are interactively getting tactile feedback, and you learn to adjust your playing and speech to get the best result. These subtle real-time adjustments are key.

It’s the same kind of subtle real-time adjustment humans make when talking... as you become aware of how your voice is carrying in the environment, you naturally adjust that volume based on real-time input (distance to target, volume/velocity necessary to cover that distance efficiently as you judge the environment). You are going to come up against this

With both the Voice source and the controlling key trigger coming from pre-recorded Tracks, gone is any ability to interactively ‘adjust’ either stream of info (easily)... but this all depends on what overall “effect” you are going for with this routing... and how this voice and control data was created.

That’s the short story. If you need more in depth details, let us know. Totally possible, thanks for the question.

Extra Credit:
I’ve looked down this hallway a few times... I had a melody recorded as a MIDI Track... I often need to demonstrate things like the Vocoder, so by pre-recording the melody I could just concentrate on moving the air molecules in front of the mic, while I explain how it works. The experience taught me is that to pull this off, the timing of the phrasing must match to achieve the best intelligibility. Your variation of also pre-recording the vocal input is a cool wrinkle... I knew it would work, but also know that the synchronicity might be important... but it may not, you’ll see as you get this dialed in on your specific data. But your theory will work, let us know how you make out...

Also, I noticed that one of the possible Modulator choices is (Master). What is this?

This is available when catastrophic feedback will not occur. If a feedback situation is attempted, you will see the source surrounded by parentheses, for example: (Master) or (Part 1). You cannot feed a Part back on itself; you are prevented by the option being unavailable (the setting will be surrounded in the parentheses). You cannot feed the Master back into system if a negative result (feedback) would be the result.

That all worked great, Bad Mister. Thanks!

In case anyone is interested, here’s my first try at doing this.

I just did this with a long lead vocal. I was worried that it might be too long, but it worked just fine.

Is there an upper time limit on the length of a “sample” like this, or is it just whether it will fit in memory?