Voice-to-MIDI Versus Audio-to-MIDI

In this article we’re going to discuss the differences between traditional Audio-to-MIDI conversion and Dubler 2’s Voice-to-MIDI.

Regarding the pitch side of Dubler, one of the most common questions we get asked is ‘Can’t DAW’s already convert audio files to MIDI files?’
The answer is yes, in that numerous DAW’s like Logic X, Ableton Live and FL Studio that have various solutions for turning a recording (for example a hummed melody idea) into a MIDI file with notes on a piano roll.

However, there are numerous drawbacks to this. The first is that, if you are not singing in key accurately, a lot of the notes will be wrong and require lots of tweaking as we will see in the examples we look at. The second issue is that Audio-to-MIDI conversion can only be done in post-processing and not in real-time. This prevents you from singing and recording live or using your voice to control MIDI during live performances. The third drawback is that you can’t use the expression of your voice to manipulate characteristics of the instrument you’re performing.

Dubler 2 is a solution to all of these and we’ll look at how in the following sections!

Two Examples of Dubler 2’s Voice-to-MIDI versus Ableton Live’s Voice-to-MIDI

We recently released a video in which we carry out the Audio-to-MIDI versus Voice-to-MIDI comparison tests and you can check that out below:

The video also contains examples of the live voice expression techniques we’re going to discuss in the next section.

Ableton Live allows you to convert recorded voice to MIDI by right-clicking the audio-file and selecting ‘Convert Melody to MIDI.’ It then creates a new MIDI instrument track with the MIDI version of the recording on it.

To test Dubler 2’s MIDI accuracy versus Ableton’s we recorded our vocals while singing guitar and synth bass parts via Dubler 2 and then converted the vocal recording to MIDI. This allowed us to have both the MIDI file created in real-time with Dubler 2 and the MIDI file from Ableton’s Audio-to-MIDI conversion.

The audio below is what we performed with a very basic guitar sound. There’s also a looping Rhodes chord sample in the background.

We converted the vocal recording to a MIDI file in Ableton and tested the resulting MIDI file on a piano instrument:

While the conversion was able to capture the timing of the notes accurately there are lots of wrong notes and even notes that are out of key.

One of the best aspects of Dubler 2’s pitch-to-MIDI conversion is the key-lock function that understands when you sing a wrong note and corrects it in real-time with inaudible latency. The result is a live recording with no notes that stray off-key.

Our second test was with a synth bass instrument:

With this one we’re making life even harder for Audio-to-MIDI conversion because it’s hard for it to pick up low notes accurately. Only a few seconds of the converted MIDI file are enough to hear that the notes are all over the place:

While performing the part with Dubler 2, however, we got a seamlessly flowing bassline and complete accuracy while hitting each note, even when singing out of key.

This isn’t to say that good results with Audio-to-MIDI are impossible but they truly rely on better singing as the algorithm is extremely sensitive.
Recording melodic ideas with Dubler can also require some piano roll clean-up but it will be significantly less. By regularly practicing with Dubler you can also easily come to a stage where no clean-up is required!

Adding Live Expression With Voice

The true power of Voice-to-MIDI lies in expression and using your voice to change the character of your instrument while singing.
Dubler 2 has many features that allow for this and we’ll provide some examples to demonstrate.

Vowel MIDI Mapping

Dubler 2’s Assign Tab allows you to MIDI map any parameter in your DAW and/or plug-ins to AAA, EEE and OOO vowels. This could for example allow you to have a filter’s cutoff open when you sing using an ‘AAA’ syllable and close when you sing with other syllables. You can also control the sensitivity of how the parameters move and the upper and lower limits for every single vowel in the Assign Tab.

For example one artist we work with, Portrait XO, is a singer and she has the vowels mapped to effects like reverb, delay and flangers while performing. Instead of using Dubler to sing melodic MIDI parts she solely uses the vowel modulation to change the characteristics of her voice. Be sure to check out the below video in which she breaks down the technique:

As another example, below we sing a pad with the Wavetable mix parameter mapped to the EEE vowel. This makes it so that the sound characteristic of the pad changes when we sing EEE vowels versus when we don’t.

Here’s one more example where we use the E vowel to open and close the filter while singing a synth bass patch:

Envelope MIDI Mapping

The ENV knob follows the loudness of your voice while you sing and, like the vowels, this can also be mapped to any parameter in your DAW and you can sett its sensitivity and upper/lower limits in the Assign Tab.

This one works great mapped to velocity parameters so we’ve mapped it to the velocity of a Sptifire Audio strings plug-in in the below example:

You can hear that as we sing more quietly (or further from the microphone) the volume decreases.

Using ENV can really help live instruments like strings sound more realistic.

Singing Chords 

The Chord Tab is one of the most exciting new features of Dubler 2. It allows you to assign chords to each note in the scale you’re using. This means that when Chord mode is enabled you can sing chords and your DAW will pick up all of the MIDI including the Chord notes.

The below audio is an example where we are using minor 7th chords in a C-minor scale. We also have the Spread voicing option enabled to spread the chord across many octaves for a wider sound:

Pitch Bend

Activating Pitch Bend in the Play tab allows you to sing and record gliding melodies. All you have to do is make sure the pitch bend range in Dubler 2 matches the pitch bend range setting of whichever synth you are using. Dubler 2 also has two pitch bend sensitivity modes that you can switch between to optimize it for your style of singing.

Here’s an example using a pitch bend range of +-12 with a lead synth sound:

It is definitely more difficult to sing with pitch bend active because you get less help from the key-lock, but it’s worth putting in the practice with it to get some great bendy melodies!

The examples we’ve looked at are just some of the ways Voice-to-MIDI can introduce a new level of performance and we’re very excited to see how people use Dubler 2 in their creative process.