Melodyne is a stand-alone program that analyzes monophonic audio — such as vocals or solo string and brass instruments — and allows changes in both pitch and time, resulting in melodic control that was previously possible only with MIDI instruments. Melodyne uses a new technology called Local Sound Synthesis, where time and pitch become completely independent of each other. So, pitch changes do not affect duration and, conversely, time changes do not affect pitch. Melodyne also automatically corrects the formant position to maintain the original sound's qualities: Men sound like men singing, and women still sound like women, even when big pitch or time shifts are implemented. Great sound quality and real-time performance are the notable features of Melodyne.
WHAT DOES IT DO?
Melodyne performs a detection pro-cess on digital audio files and generates a Melodyne Definition Data file that must reside along with the audio file in the same folder. This file, with an .mdd extension, fully describes the individual pitches, including unique vibrato of all notes; timing, including note lengths and internal tempos; pitch transition times, or the time it takes to go from note to note; individual note amplitudes (volume); and formant information (resonant volume). Any of these parameters and their interaction are completely editable in a highly intuitive visual interface, with real-time changes possible on audio streams of up to 24 separate tracks playing at a time.
No “rendering” is necessary for the user to hear results; changes are totally elastic until they are saved to the .mdd file. The original audio files are never altered, and, once all changes are made, a version can be saved as a new audio file and placed back in your session, whether you are working in Pro Tools, another DAW system or with a good ol' analog tape recording. There are several output options, including a stereo mix of melodies (as in an ensemble) or separate melody tracks (say, for separate lead vocal and harmony tracks). You can also output standard MIDI files that play the detected melody in exact sync with the audio. With this first released version, the exported MIDI file contains no dynamic pitch change information, such as any vibrato or glissando/portamento a singer may use; the MIDI file only contains the note number and note on/off. Melodyne accomplishes this digitally divine handiwork of interpreting your ideas in a natural, musically intelligent way, and without altering the basic nature of the sound file.
INSTALL AND GO
Installation went fine from the single CD-ROM on our G4/400 Mac with 704 MB of RAM running OS 9.1. Minimum system requirements are: Power Mac G4 or G3 (blue and white), PowerBook G4, PowerBook G3 (Wallstreet, Pismo), iMac, iBook and 128MB RAM. MAC OS 9.0.4 or later (including MAC OS X) is required. If you don't have CardbonLib 1.3.1, then you'll have to download it from Apple's server or from Celemony's Website: While online, you'll get an authorization (Celemony) that takes about three minutes. The manual takes you through many tutorial examples to learn this deep program using the supplied undetected and detected melodies. Melodyne runs under either the Mac's Sound Manager or any ASIO driver. ASIO allows more audio channels, lower latency and other sample rates besides 44.1 kHz. For this review, I mostly used Sound Manager, which has just two outputs. Melodyne has a basic, onboard mixer page for mixing down to stereo or assigning multiple tracks to separate outputs when saving processed sound files.
DETECTION AND ANALYSIS
Melodyne works with .AIFF, SD2, .WAV and .SND sound file formats. Sounds must be totally monophonic-only single notes at a time, dry without effects. I was unable to use interlaced stereo files, but you can import separate left and right channel files. Detection of a new sound file happens as soon as you import it into Melodyne's Arrangement window and double-click on it. There is a Melody Definition window for predefining what the program expects to “see” and for correcting detection mistakes. Even though the program rarely misdetected, this feature is important for melodies for which you already have tempo and scale information and want to make sure Melodyne gets it right. You also use this window when importing a reference backing track that you have no intention of processing. If you know nothing about the new melody, then the program will detect melody tempo and pitch, and it will attempt to detect the scale used within the melody.
The Arrangement window is the main window where all melodies (audio files) are shown in color-coded organization. There are transport controls with a Looping function, individual track record arming and naming, and Solo and Mute buttons just as there are with any other DAW system. You can also get an instant, notated score view of the melody if you prefer to work that way. At the top of the screen is a grid timeline indicating bars and beats; however, in this version, only quarter-notes are shown, without subdivisions or ticks for precise locating. Time-stamp information of audio files (such as Pro Tools or Logic Audio files) is not recognized in the program, so exact synchronization depends upon all audio files starting at the same time when they are imported into Melodyne. I am told that time-stamping will be supported in later versions. For now, there is a Bar Quant feature that will “snap” any new file to the nearest bar (as selected in the Quantizing menu) that facilitates lining up audio files.
There are icon/tool sets that are contextually based on the setting of an Action pull-down menu and what operation you want to perform. There are also digital readouts for detected tempo, time signature and quantize. Depending on the Action submenu, you can zoom, play individual notes, “scrub” or solo. The Arrangement window is where you can record or import sound files; rearrange by copying and pasting single notes or whole melodies; define, adapt or change the tempo of the entire melody; and slip an entire melody in time. To edit individual notes, you must double-click on the melody and use the Editor window.
The Editor is for already-detected melodies, and if you double-click on an undetected melody, it will undergo detection and open. This detection process is amazing! Instead of displaying a straight waveform from left to right across the screen, Melodyne passes the entire waveform into “chunks” vertically up and down, corresponding to the individual notes detected, in-line with a piano keyboard on the screen's left side. Because each distinct note is a separate event, the internal vibrato of the note, as well as the transition from note to note, is graphically represented along with each note's amplitude and timing. This is where the fun starts!
Depending on which Action submenu and which tool you select, the melody is marked up with color-coded glyphs that show which melody parameters are available for adjustment by point, click and drag. Without naming all of the different tools, you can create new melodies with copy and paste or replace single notes; freely move notes in time; alter time phrasing; or quantize to a selected grid depth. I found this excellent for lining up a poor double-tracked vocal to a well-sung lead vocal, and for stretching or shortening certain notes for artistic reasons (slowing down a note where it literally stands still and have it sound like the singer held it without ever running out of breath!). You can correct intonation of individual note events or slice note events into smaller bits for microscopic fixes (like having Auto-Tune built-in).
There is also the ability to quantize or align a melody to a defined scale. Melodyne will detect the scale, but you can change it from melody to melody if you wish. Melodyne, when pitch correcting, will not alter vibratos, and you can separately manipulate vibrato intensities all the way to flattened-out for that “CherBot” sound. I guarantee that changing the speed of pitch transition from note to note will become a favorite thing (just like Auto-Tuning) with young singers!
You can change formant position of individual notes or entire tracks. This is gender-bender stuff, but I found it easy to change a trumpet to a trombone or a tenor male singer to a high baritone. I also could mute or alter the amplitude of any note and then modify the attack portion of a note. This is more useful for percussion processing, but has definite potential for certain vocal sounds.
Melodyne has specific playback algorithms especially tailored to different sound sources. The different algorithms use varying amounts of available processing power, depending on the complexity of the sounds. This becomes important, especially with many tracks playing and minimal computer resources. There are five different algorithms placed here according to required computational power: Sampler, Time, Pitch, Formant and Voice. The least resource-hogging are Sampler and Time, which are used for undetected melodies. Sampler is just like a simple sampler playing backing tracks, and Time is optimized for percussive material in which each attack is carefully preserved. Pitch applies Local Sound Synthesis to the entire sound file and is suited for vocals and instruments with soft transitions between notes. Formant is when only the formants are changed, and Voice is for changing both pitch and formant in correlation. To conserve computer resources, Melodyne automatically determines which playback algorithm to use, depending on what parameters you have manipulated. Playback algorithm is applied on a track-by-track basis, so you should not put, for example, a percussive sound on the same track as a vocal.
Even with this early Version 1.0, I am enthusiastic about Melodyne because of its huge potential and immediate usefulness. But Celemony's seemingly Faustian bargain comes with a few caveats that are promised to be fixed in Version 1.1. First of all, there is no undo. There is a Reset button that resets the entire file back to the original detected state if you are in the Arrangement screen. A simple Command-Z to undo the last operation, like in most Mac programs, would be a big help! Furthermore, if you delete a note from a melody, there is no way to get it back other than to re-import the file…pretty lame. By the time you read this, Multiple Undo will be available.
I found four major bugs. 1) The Apple Menu icon dims while Melodyne is present, so you have to click on the desktop to get the Apple Menu back. 2) Melodyne has trouble knowing the difference between closely similar file names when importing. 3) The Playback algorithm function disappears mysteriously from time to time but returns on reboot. 4) While using either an ASIO driver or Sound Manager, I had a 10dB lower level when exporting processed files back to Pro Tools. Carsten Gehle and Joerg Huettner at Celemony explained that later versions will fix all these shortcomings. Also promised for future versions, Melodyne will synchronize with other programs and redirect its output into their mixers; once connected, any changes of the tempo of your sequencer or hard disk recorder will immediately be reflected by your audio. Also, specific versions of Melodyne will be available for both VST2 and MAS-based front-end applications (Logic/Cubase/Digital Performer), as well as Pro Tools once the legalities with Digidesign are resolved. Other plans for V. 1.1: configurable keystroke command shortcuts; revert to save; stereo interlace file support; and more intensive MIDI performance.
I am thrilled with how well Melodyne works and sounds, even though I did experience a few freezes, crashes and Type 2 errors along the way. I look forward to future updates and revisions when Melodyne becomes a world-class, fully professional piece of software.
Melodyne sells for $995 MSRP from Celemony Software GmbH, Valley Strasse 25, Munich, Germany 81371. In this country, contact the GSF Agency at 118½ Pacific Street, Santa Monica, CA 90405; 310/452-6216; fax 310/452-3886; firstname.lastname@example.org. Or visit www.celemony.com, where you can download MP3 files of processing examples. Thanks go to producer David Gamson for the use of his studio/computers and helping me with my evaluation.
Barry Rudolph is an L.A.-based recording engineer. Visit his Website at:www.barryrudolph.com.