Where are your levels?

[Trey LaCroix CAS]

I just had a conversation with the post sup and editorial (post sound isn’t hired yet) for an upcoming movie I'm mixing. When the subject of levels came up they asked how I typically record. I told them my standard answer of “I aim for as hot as I can without hitting the limiters except on very dynamic scenes where I record a bit more conservatively.” They replied with “we need tracks to be hot all the time so we don’t deal with any noise issues.” Putting aside the probable past bad experience they had with other mixers (which is the most likely scenario to me) I think it brings up a good point. Why are we still recording so hot? This made sense back in the day of getting over tape hiss or 16 bit dithering noise but I really question the validity of it in a modern, fully digital workflow.

 

The mixers who trained and mentored me early on got me in the habit of recording "hot" typically peaking around -6ish dbfs on my mix and maybe a tad lower on my isos. Post has typically never had issue with this as long as I don't slam the limiters and am quick with my gain knobs. Editors have even praised told many times that they are grateful for my hot tracks. However, in recent years I have moved away from my Lectros  (and their softer limiters) to a fully digital workflow using 100% digital wireless (Shure Axient and Zaxcom) and signal path (Dante or AES) and started recording more conservatively. I can capture the dynamics of the scene without fearing a sudden performance change that wasn’t in the previous take to slam the limiter and, if my understanding of PCM audio is correct, know that the noise floor of my 24 bit equipment is hardly a factor. As long as the preamp and A/D of the wireless is properly driven the S/N ratio has been established and it’s all 1s and 0s that can be arbitrarily raised and lowered from that point on. I still very much keep levels healthy and still fairly hot if I’m able to, but I let things breath a lot more now and I believe (and have been told by several post mixers) that the dynamics make for a better product on their end. I get that it’s nice to have full, loud levels in post that need minimal gain adjustments, but sometimes it seems like  it is at the detriment of the actual product. I can’t tell you the last show I did where we had proper rehearsals or consistent performances from actors.

 

I’m curious what people think and if anyone else has tracked more conservatively over the years? Or, if you think I’m full of crap I’d love to hear that too!

[drpro]

+5 all the way!!

[humbuk]

I try to do the mixdown for the editing room as your "standard answer" is. In the editing room they don't want to deal with the sound, so if they have it at full volume, they are satisfied :). And of course the ISO tracks at such a level that I don't have to move the gain too much during the shot.
After all, in the mixdown for the editing room, it is mainly about having a balanced volume and the right proportions between the individual tracks (microphones, actors), but if the sound sometimes leans on the limiter, it is not so important, in my opinion.

[fieldsound]

I use wisycom wireless and have my gain staging set up so the limiters basically never engage on the transmitter side. This is in contrast to how I used to work with lectros, recording a lot hotter at the transmitter to get solid modulation on my receiver

 

I used to gain the mix really conservatively and now I go a bit hotter on my 8 series. A lot of stuff I work on will use my mix for broadcast so I tend to lean a bit on the hotter side for that type of thing. I don’t mind tickling the limiters with my peaks at all. 

 

That being said if I’m working on a narrative project that will have more robust post production I do try to leave more room for the mix to breathe and will sometimes record a safety track for extremely dynamic scenes with screaming etc. Truthfully, our signal chains are incredibly quiet compared to most of the bg sounds we regularly contend with shooting on location. 

 

[Philip Perkins]

One of the best reasons for split prefade-iso and location mix recording is that you can leave the isos fairly low (if you think there are surprises in store) and go wild with your mix--even using the onboard bus limiter to compress it some.  I never had anyone complain about compression in dailies mixes--the more compressed I made them the better they liked them.

[JonG]

In my experience, when post requests this, it’s a lack of experience on their part. 

[The Documentary Sound Guy]

11 hours ago, Trey LaCroix CAS said:

Why are we still recording so hot?

 

I've been wondering this for years.  I aim for average levels at -20dBFS, but don't worry too much about it otherwise.  This keeps it in the ballpark of the -24LUKS that many productions are mixed to.

I do get the odd complaints about mixing too quiet and I try to oblige when asked, but I do find that I end up hitting the limiters more than I would like if I'm much above this level.  Usually I get this complaint from directors / indie filmmakers who are checking the raw files on a laptop or something with weak speakers that are designed to playback consumer music or some such.  I don't recall getting the complaint from post.

[Philip Perkins]

I can tell you that underrecording often results in picture editors doing the dreaded "clip stacking" to get the volume up.  By this I mean that instead of dealing with clip levels etc they just copy and paste multiple copies of the same clip on the tracks below the original and play them together to bring the level up.  On quiet scenes I have gotten as many as 5 copies of the same clip stacked up to bring the volume to where they wanted it.  I won't get into how lazy and inefficient this practice is but it is the result of too-low dailies mixes.

[Derek H]

 I would think we should be recording the deliverables that the dialog editor and re-recording mixer need and not the picture editor. It’s hard to please everyone.  I’ve received the email before about “low levels” and have had to bite my tongue and not point out how nice it must be to have the screamed line clean that followed the whispers. I think from a post perspective they don’t always understand the challenge of recording actors in a dynamic scene in real time. 

[humbuk]

11 hours ago, Philip Perkins said:

Můžu vám říct, že podhodnocení nahrávání často vede k tomu, že editory obrázků provádějí obávané „vrstvení klipů“, aby zvýšily hlasitost. Tím myslím, že místo toho, aby se zabývali úrovněmi klipů atd., jednoduše zkopírují a vloží více kopií stejného klipu na stopy pod originálem a přehrají je společně, aby zvýšily hlasitost. U tichých scén se mi stalo, že jsem naskládal až 5 kopií stejného klipu, aby dosáhl požadované hlasitosti. Nebudu zacházet do toho, jak je tento postup líný a neefektivní, ale je to výsledek příliš nízkých denních mixů.

Yes, that's exactly what I encounter quite often. That's why I mixdown as loudly as possible. There's no point in speculating about how competent the editors are and how easy it is to amplify the sound in an editing program, but that's just the practice and it's easier to adapt to it.

[igomarsound]

interessant disccussion.

 

for fdialogue I try to navigate around -20 DBFS, on the quiet parts, -12DBFS and max -6 DBFS on my iso tracks, it's all about headroom and healthy signals, but it's already pretty hot in my opinion but i also got some complaints from post before so here's the compromise i managed to live with.

 

my mix, OTOH, is way more hotter cranked up between -6 and -3 DBFS, even hitting the limiters sometimes because thats what the editors will use and they don't have a clue and just want it to sound BIG and carry on with their workflow. so here. also while i send the LR mix to directors and cameras, they ll get this medicine too. not a fan of this gain staging chain but hey that's what people want. i provide, 

 

it would defintely make more sense to have the loudness approach, target at-23LUFS, but i dont see that coming.

[Roland Harris]

Well, the gear you use comes into play heavily. I've just been dipping my toe in the water of 32-bit recording, with a Tascam FR-AV2 (modest, but seemingly has the same preamps as the very nice sounding HS-P82), and as one of the tests of this new device I have today run a series of tests on self-noise comparing this to a 788T and MixPre-3 (1st gen: 24 bit) at max gain (both at 76dB), reducing gain in steps (56dB, 36dB, 16dB and 6dB - the last being min gain on the MixPre-3), but adding gain in post to bring the level back up to 76dB. All three recorders have about the same EIN, so have essentially identical self-noise at reasonable gains, but, as you would expect, by the time I got down to 16dB gain (with +60dB in post) the 788T was a little hissy; and at 6dB gain (with +70dB in post) it was very hissy. But what might be less expected for some, is that the MixPre-3 had no more self-noise at its min gain of 6dB (with a +70dB boost in post) than at 76dB gain with no boost in post. Obviously different 24-bit preamps/recorders behave differently, but that is fundamental to the 'how hot should I record?' reality (assuming post are happy with low levels!): what applies to a 788T does not apply to a MixPre mk 1 recorder etc. etc.

Cheers,

Roland

[The Documentary Sound Guy]

Based on that, I'd be willing to bet that 'gain' on the MixPre is actually digital gain, not pre-amp gain.

[Roland Harris]

11 hours ago, The Documentary Sound Guy said:

Based on that, I'd be willing to bet that 'gain' on the MixPre is actually digital gain, not pre-amp gain.

 

You'd lose the bet, I'm afraid. Putting aside that this would make no sense of the results (!), that consistency in EIN and dynamic range is quite in-keeping with the Kashmir preamp's topology. You can see a tear-down showing this here should you be interested. But my simple point was that all preamps/ADCs are not the same, and that self-noise at extremely conservative gains will be possible on some and not others. Although, of course, post may not thank you when handing over such files: that's a very different and real issue, and not entirely removed from the workflow issues arising from 32-bit float.

 

Cheers,

 

Roland

[The Documentary Sound Guy]

Thanks!  That tear-down looks super interesting.

 

[codyman]

On 10/8/2025 at 10:33 AM, Roland Harris said:

by the time I got down to 16dB gain (with +60dB in post) the 788T was a little hissy; and at 6dB gain (with +70dB in post) it was very hissy.

I think you might be the first person ever to complain about the preamp in a 788t.  It's like the Lexus LS400 of preamps - it's got years on it but still keeps driving past most of the competition.

[Roland Harris]

Not remotely complaining about the 788T: I have two, and have no intention of ever letting them go as I think they are wonderful! They just have a different preamp and ADC topology/design, which is what SD have always said. 

 

Cheers,

 

Roland