Record volumes when using any MIDI and Audio based system requires a bit of understanding about the items in play. For example, if you record MIDI data first, this is often done with no relationship to what it will result in when converted to audio. A quick example, (and it is a natural thing for all of us to do as musicians)... but say you record Drums, Bass, Guitar, a couple of Keyboards sounds (ep, organ) and then you go to add some strings as sweetening in the background. You will probably do what is akin to pre-mix the strings. Because they will not be up front in your final mix, you wind up recording them as MIDI at very low velocities. You may look at your string part and see that you never hit the keys with a velocity of more than 40... This, of course, sound fine as MIDI ... but

Later when you go to transfer it to audio, you discover that the record level is "too low". Can you see how this can happen? It's because when we record MIDI we are "musicians", and when you transfer it to audio you become an "audio engineer". The audio engineer, being a technician, has requirements of levels and making meters move; while the musician, as the artiste, is unaware of the technical requirements and just played it to make it work musically.

Had you recorded the string part as Audio, you probably would have set your record level using a METER (rather than just your artistic sensibility), and you would have set the LEVEL much, much higher, then used a Fader to mix the volume down so that it artistically fit.

You can either start thinking "technically" a bit earlier in the process (experience will teach you this) or you can learn a few tricks that can help you adjust the OUTPUT LEVEL of your recorded parts even after the fact... this way you can continue to work as you naturally would, but can prep the data for recording as audio.

Below is taken from an article I wrote on MIDI Recording and Audio Record Levels _ sorry for the length but it is all right on point to your question -especially Point #2 - which explains the difference between the VOICE volume and the PART volume... (hope it helps):
.........................................................................................................................................................................

Ways to adjust PART VOLUME for audio LEVEL
1 _ Raise/lower the CS for that PART (PART VOLUME)
2 _ EDIT the Voice’s Volume (this can be accomplished while in MIXING mode via the VCE EDIT function)
3 _ Work with the PART Velocity (with knowledge of how it may affect performance parameters).
4 _ Increase mLAN/FW OUTPUT level

1-Raise/lower the individual CS (Control Slider) for that PART
Use the CS to affect output level of each individual PART. Reference the DAW’s ‘mixer’ so that you can determine the overall level (with the METER). The object is to get a good LEVEL yet without any clip indications. A clip would be any positive value in the meter’s PEAK indicator or red “CLIP” indication. Keep you ears on the music and your eyes on your meters! Get good LEVEL with NO CLIPS!!!

2-EDIT the Voice’s Volume
Press [MIXING] > [F6] VCE EDIT - this will allow you to drop into full Voice Edit on that sound. This will let you edit, where necessary, the Voice’s overall VOLUME parameter. Here’s how:

From the MIXING screen
_ Press [F6] VCE EDIT
_ Press [COMMON EDIT]
_ Press [F1] GENERAL
_ Press [SF2] PLAY MODE
_ Adjust the OUTPUT LEVEL Volume parameter, as necessary

You will find many of the Motif XF Voices are very conservatively set for overall Volume… so you have plenty of room to move the output level up or down. This would be like turning the instrument up "at the source". If your MIXING screen is like a studio’s mixing console, and the fader on that mixer is just like the fader on a console, then this OUTPUT LEVEL Volume is the equivalent to asking the musician to give you more input level (from the source). Knowing how to turn the instrument up at the source is a fundamental that you need to know about. You will find that Yamaha has most often left you plenty of room to turn things up louder - and this is by design. If a VOICE is not giving you enough level with the FADER alone, ask the individual (instrument) to give you more gain (at the source). This OUTPUT LEVEL Volume is this gain "at the source!"

If you are getting low levels on your METER in the DAW ‘mixer’, remember as ‘audio engineer’ your job is to get the best-recorded level on the media without destroying the musician’s performance. So the engineer’s job is strictly one of getting the sound recorded properly.

Once you have edited the Output Level Volume of the PART, you can store this edit to a MIX VOICE location provided for each normal Voice. Each Drum Kit Voice that you edit by this method must be stored in the [DR USER] bank (Drum Kits cannot be stored locally within the SONG/PATTERN)

_ Press [STORE] to the keep the results in a ‘local’ MIX VOICE location.
Press [ENTER] to execute.

By storing the Voice in the MIX VOICE bank, the Voice becomes a part of the local SONG MIXING or PATTERN MIXING setup. This special MIX VOICE location (there are 16 of them per MIX) is provided for all normal (that is non-drum, non-track sample) Voices. It is provided specifically so that you can edit a Voice and tailor it specifically for this particular mix. Take advantage of this memory location – it allows for all your tweaks to Voices for this Project without having to change the original source Voice. Drum Kit Voices must be stored to one of the 32 User Kit locations.

The XF can hold a maximum of 256 MIXING VOICES (32 User Drum Kits).

You will find that most of the Voices in the Motif XF have very conservative Output Level volumes – this is done purposefully so that you can ‘mix’. Voices are designed to be used in combination. For example, when you are in Performance, or in Song/Pattern Mixing modes it would be quite easy to overload the input of a connected device if every VOICE was set at MAXIMUM. If this is difficult to understand - think about it for a minute. Say YAMAHA had maximized the volume of each Voice to 127, and then you attempted to layer Piano, Strings and Brass in a PERFORMANCE. You’d hit the first note or chord and you’d distort your mixer. When working in a PERFORMANCE you are combine as many as four Voices, however, when working with a sequencer you are combining as many as 16 Voices!

The Signal-to-Noise specification is so good in digital gear these days that you can use this “headroom” to your sonic advantage. So do not stress over this - if you can hear the noise floor in the Motif XS/XF you must have superhuman hearing.

A clip is a reason to do something over – there are no reasons to ever clip a signal. Every clip diminishes the total quality of your finished product. (I cannot repeat this enough)!

3-Work with the PART Velocity
Changing the velocity of the performance should be your last resort. As mentioned above, it does have much to do with the volume that you are able to achieve from each individual PART. Many times no attention is paid to how much velocity is used when recording data as MIDI tracks and you must compensate for extremely low output from track data. Fortunately, there is plenty of room to edit XS/XF Voice Volume data. However, the warning about changing the velocity of a performance is only given for those particular VOICES that change character when velocity is altered. Velocity can be safely offset on many Voices in the Motif XF.

The reason for a warning when adjusting velocity is simply this: If you have a VOICE that is a velocity swap - this means it changes character by switching to a different sample for a completely different articulation - then you want to be very careful when altering the musical performance velocity. A good example of this is a Voice like a SLAP BASS… when you exceed a particular Velocity the articulation switches from a normally plucked attack to a ‘slapped’ attack. ‘When’ that occurs can be altered if you manipulate the VELOCITY parameters - just be aware of how the sound and performance change if you decide this is the method you wish to use. You can consult the DATA LIST Booklet - the WAVEFORM LIST will indicate Voices that have multiple velocity articulations.

One way to adjust the volume is by artful use of the VELOCITY SENSITIVITY DEPTH and OFFSET parameters. These are found from the main SONG or main PATTERN screen:

Press [MIXING]
Press [EDIT]
Press the Track Select button of the PART you want to offset, [1]-[16]
Press [F1] VOICE
Press [SF3] OTHER

The two parameters: Velocity Sensitivity Depth and Offset, allow you to change the velocity of data on a Motif XS track. An Offset adds or subtracts a value to each velocity value. “64” is no change.

4- Individual mLAN/FW OUTPUT boost
You may notice that when you route PARTS to individual outputs you now want each PART to reach a greater gain level on the meter. We can add this fourth method of increasing the output level of the individual PARTS. This is necessary because when you recorded your tracks initially you (probably) were not paying attention to the audio footprint they would leave. You played the string sound softly because that made sense while recording MIDI, but recording audio you now want that same performance but you are seeking to optimize the record level. You are basically undoing the mix balance you created in order to record the individual audio tracks. Your goal is to later recreate a mix balance with these musical performances as audio tracks.

Press [UTILITY]
Press [F2] I/O

Here you see you can increase the output level (GAIN) of the individual mLAN buses. If you require more level (and most likely you will) you can get an additional +6dB boost for each mLAN pair. The DAW MIXER can usually add an additional +6dB of gain as well (but try and avoid using the computer to do this - as it will undoubtedly add noise). Do not stress if your individual PARTS to not reach 0dB – your goal is to get a good sounding, clean recording of each Part so that when you do your final mixdown everything can be heard. That is it. You get no extra points for reaching 0dB on each PART. Quality of the Sound wins!

We cannot conclude this article without mentioning that learning to use a Compressor can make a huge difference in your overall mix output. In the Master Effects area, you can select from the VCM COMPRESSOR 376 and the MULTI-BAND COMPRESSOR. A compressor is a leveling amplifier that is typically used to finalize a mix. It can increase what we commonly call the “presence” of a mix - it can put the mix “in your face”. Unfortunately, it can be over used - but learning and experimenting with these, particularly when transferring a stereo mix out from the Motif XF, can be its own reward.

In short, a compressor allows you to reduce the dynamic range (it allow you to bring the soft closer to the loud). Artfully done this can make a world of difference in your final result. And will be the topic of separate articles. Enjoy!
Hope that helps…