I'm mixing a video voice over. One voice track, one music bed track. I'm manually ducking the music using volume automation at moments of speech.

The voice track has a fairly constant level. But the music track slowly increases in volume to a minor crescendo at the end. Not a lot - a few db, but enough that the constant-level voice becomes lost in it.

What's the standard thing to do here? Hold the voice level constant and duck the music an increasing amount at moments of speech? Or increase the level of the voice and apply a constant amount of ducking to the music?

Thanks.

Unless this show is about the music, dialog should be king. Which means you do whatever eq and level adjusting is needed to make the voice sound natural. In the case of a narrated video, it can also be acceptable to add some level compression (but beware of the factory presets in most compressors, which are designed for music).

Once dialog is right, then mix in the music. Duck it when necessary to make the voice more important and follow the message. If the music goes up during a narration, ride it back down. Or edit the music, so the exciting part is in the clear...

Hey, thanks so much for the clear direction. Of course, I hadn't appreciated the importance of this detail - yes, it's in fact an instrumental music video, and I'm over-dubbing a commentary. There's only five comments, about 10 seconds each, in a 3 minute video. You know - promo stuff - "Introducing, Band xyz"..da da da da....."check them out on your favorite streaming service today".

I agree. The only thing I would add is to check your mix in mono.

Honestly I think you'll have to use your own judgement here. It needs to be both exciting so that we want to check out this band on this streaming service but we also don't want the narration to yell at us in the middle of a song. So you need to probably find a nice middle ground there.

I was thinking that exact thing this morning, thanks. Maybe I'll bump the voice up a dB or two overall, then duck the louder music in the second half an extra dB or two. Let the voice remain constant, set the level at the outset for the listener, yet avoid too much pumping the music towards the end.

Sometimes it helps ducking music in only those frequency bands which conflict with the voice.

We are on the same wavelength there :-) I've been experimenting with that. Sometimes it sounds really good, making the voice sit nicely in the mix. Sometimes the change in EQ on the music makes some elements go odd - like leaving too much high end sizzle on strings momentarily 'cause I've ducked the voice mid-range.

You might already be doing this, but if not:

Try putting a multiband compressor on the music track and feed the voice over into the side chain input of the compressor. Set compressor to duck the mid to high-mid bands by 3-5dB or so and see if that helps.

That's essentially what I've been experimenting with, but using Fabfilter Pro Q3 EQ instead of a compressor. I set an eq point, then automate the gain reduction at moments of speech. Gives me complete control over frequency, Q and gain.

Yes. And react, don't anticipate. The ear always hears the "new thing," so it'll pick out the starts of the phrases. Be mindful, though, of the ends of phrases -- don't let them fall off the table, and ride the VO up in those spots rather than pulling the music back.

Actually, a few ms of lookahead dont do any harm..

I'm personally not a fan of duckers and "carving" and all that. But I probably hear stuff that's been done that way all the time and don't notice.

Well, I'm using a sinusoidal-shaped curve (that is, from max to min and back) to automate the level duck. And I've been placing the level-drop so that it reaches bottom just as the voice starts. You're saying I should place it to start the drop at the moment the voice starts?

If I'm saying anything, it's that I have no need for a ducker. But if I were forced to use one, I wouldn't have it ducking in advance.

So, clearly I have to ask about this. You mix voice over music without ducking the music at moments of speech?

By "a ducker" I mean an automatic device. I do the ducking with my finger.

Ah, I see. So we're doing basically the same thing, only I'm doing it in automation. Thanks.

If it's not a live mix, I use automation too.

The main point, to me, is that the listener should never hear you "mixing." If you're sneaking the music down in anticipation of the VO coming in, that change can be a distraction. The music might not be fighting the voice, but the changing music level might draw away the listener's attention, in which case you've effed up.

This ^

I have seen/heard a few trailers (recently) where the music is moving along nicely, and as soon as dial/narr enters, the music takes a tumble of more than a few db. I, in my mind's eye, could actually see the fader being jerked downward. It is in fact distracting, and jarring as well.

The problem a lot of the time, is it's just the wrong piece of music. The music is supposed to be composed/arranged with the dial in mind, so as to let the dial be heard clearly, cleanly and, as said earlier, to be king. But much of the time, music is chosen from a batch that was just written/arranged/produced. The composer has no idea when/where/how it's going to be used; he/she just writes and hopes for a placement.

op: have you tried any techniques to auto-duck the freqs of the music, based on where the dial sits? I have never tried anything like this, so I have no idea how it might work... just a thought.

Cheers.

I'm not sure I agree. I mean, this seems to be a bit similar to any time someone says that "too much" of something is bad, which, well of course it is... 'cause that's what "too much" means.

I don't really have a problem with music coming down before dialog enters, unless I have a problem with it. Know what I mean?

I really think this is a case where the engineer just has to have a musical ear and be sensitive to it all. It makes a huge difference if there's a hit slightly before dialog comes in or if it's just pads... or if it's full range music or pretty sparse in the range where dialog sits.. and is the incoming dialog loud and forceful or are we dealing with an emotional moment in a documentary..

Completely situation dependent in my opinion and I often drop level before dialog starts with zero complaints from clients as long as I do it "right".

Doing it right is the main thing. I beat myself up constantly about things the client didn't complain about.

Yeah, well, I suppose I think I'm doing it right and the client isn't complaining... in case that wasn't clear... :-/

All good points. You should also have a finger on the dialog volume, making subtle adjustments on syllables... not to flatten out the volume, which a compressor could do and which can hurt the performance, but to help the performance with tiny helps for syllables that should be stressed or un-stressed.

Assumed in all this is that you have a knob and fingers available. Doing it with rubber bands in an NLE is virtually impossible... there has to be realtime responsiveness between volume control and dialog while you're listening. Doing it with an on-screen fader and mouse is incredibly hard, both because of latency and because the best mixes use multiple fingers, often in different directions in realtime. That means a single-fader USB add-on is a poor compromise.

MIDI controllers are pretty cheap. You can get a used Behringer with 8 motorized faders on eBay for ~$90, and it's good enough to make a big difference in mixing compared to any on-screen solution.

^^^^^ ... times a million. And not just for Post. Music, too.

No, what Brent is saying is, use your finger on a fader, and lower the music, don't anticipate the voice over rect to it.
Meaning don't duck the music BEFORE the voice over, but just as it starts.

Hand/ear coordination works best.

Sometimes you want to hear the ducking, and sometimes you don't.
Depends on the style of the material and intentions.

A hint of ducking few hundred ms before the VO kicks in can be good to give the listener a heads up, especially if the VO is as sparse as you describe.
But again, this all depends on the VO style. If the material and the VO are amped up and high energy, then I'd usually try not to let the density of the mix drop by ducking too early or too much.

Well, I applied several of the concepts in this thread and am reasonably happy with the result.

I tried experimenting with EQing the vocal mid-range with gain automation instead of volume ducking, but the voice is a deep-ish male voice with a lot of mid and bass and the EQing required to not conflict with the voice was excessive - or if I EQd the music reasonably the combination of music and voice mid-lows created a lot of mud. I took out a little bit of the voice's lows, but it was important not to lose the rich timbre of the voice.

So, I ended up using just good old volume ducking, using automation on the fader in the DAW.

I butted the volume drop right up to the voice starts, and made it somewhat rapid, bet made the volume-ups a little more gradual.

I set the voice volume to a constant level, then rode the voice volume fader - in automation again - just a bit in spots to avoid the ends of phrases tailing off and getting lost. But I did not overall increase the voice volume as the music increased. Instead, I increased the music ducking.

Then adjust the depth of the music ducks so that the voice was clearly audible without jumping out.

Overall, a satisfactory result.

Thanks to everyone in the thread for your time & help.