Audio Clip Normalize Including Effects Chain

In DaVinci Resolve 18.6.2 and the public beta release 19: When an audio clip is normalized, the procedure includes the effects chain for the clip processing in the assessment. I don't know if anything else is included.

1. Firstly, normalize the clip without any effects or other processing- note the audio levels and the peak display.
2. Place a gain-changing plugin, such as a compressor, into the effects chain.
3. Set the plugin to significantly reduce the gain.
4. Run the normalize function again.
5. Disable the plugin.
6. Compare the audio clip levels now to what was before.

You will find the clip gain much higher than before, roughly or exactly equivalent to the gain change in the effects chain.

The effects chain should not be included in the audio clip normalize function. The function should set the input audio level relative to only the audio clip before further processing. This is an industry standard.

To include the effects chain in assessing the level and then applying that back to the clip introduces mayhem.

You can see the result at the input volume for the clip processing.

What is now happening is a bug. Please fix this.

The effects chain should not be included in the audio clip normalize function.

I disagree. If you don't want effects included, Normalize before adding them.

The effects chain should not be included in the audio clip normalize function.

I disagree. If you don't want effects included, Normalize before adding them.

The only reason I could see someone saying this is that they're after a "post clip and processing" normalization.

Otherwise I stand my ground along with all of the rest of the industry.

Clip Gain is pre-Fairlight Track FX, Inserts and Fairlight Equaliser and Dynamics.

The Clip Normalise function is non-destructive and the same as manually adjusting the Inspector's “Volume” slider or the Gain line on the Clip itself - but providing an automatic method to conform the level of source Clips.

DSP processes in the path will result in the source no longer conforming to the required target level. The Normalise function is auto-conforming the source Clip to a specified target.

The Normalise function can also be useful, for example, to automatically level match source Clips with DSP to the source level without manually adjusting Clip Gain. If additional Gain is required, either manually adjust Clip Gain (which is pre-DSP) or use the Path's Trim control (which is post-Fairlight Track FX, but pre-inserts and Fairlight Equaliser and Dynamics).

Clip Gain is pre-Fairlight Track FX, Inserts and Fairlight Equaliser and Dynamics.

The Clip Normalise function is non-destructive and the same as manually adjusting the Inspector's “Volume” slider or the Gain line on the Clip itself - but providing an automatic method to conform the level of source Clips.

"pre-Fairlight" but not pre-FX? It would be fabulous if it kept the full context of source level assessment and source level normalization before presentation to the rest of the processing without including anything else whatsoever.

I'm looking at the clip processing as a mini-channel strip. I need input source normalization to be input source normalization regardless of what happens in the rest of that mini-channel strip. This is the only implementation I've come across that doesn't.

The current implementation adjusts the level back at the source attempting to affect the level at some point in the signal processing which at this point includes the FX chain. This is no longer normalization. I don't know what you call that.

DSP processes in the path will result in the source no longer conforming to the required target level. The Normalise function is auto-conforming the source Clip to a specified target.

I would agree. That is no longer normalization in the conventional sense. I didn't add the FX into the normalization algorithm. I would like that fixed.

For folks needing the output from the clip processing to be normalized, that should be an additional function. Please. Totally separated from source normalization.

It may be clearer if the “Normalise Audio Level” function was named “Conform Audio Level” as it is not a normalise function but rather one that conforms the source to a target level.

It may be clearer if the “Normalise Audio Level” function was named “Conform Audio Level” as it is not a normalise function but rather one that conforms the source to a target level.

A target level at some point in the processing chain...

I've had difficulty finding any audio flow charts laying out the processing. Those can be convenient. They can help avoid difficult discovery moments. Apparently, the Clip Processing is one of those places.

I would like a true source normalization to be put in.

Leave what you call Conforming as is. One other person likes it specifically that way and I have no information as to why. It is very unconventional, this Conform, as you call it, procedure.

Conforming means something else to me.

Conform to or comply with a set standard or guidelines.

It may be clearer if the “Normalise Audio Level” function was named “Conform Audio Level” as it is not a normalise function

What is audio normalization?

In the simplest sense, audio normalization is the process of measuring some aspect of an audio file and adjusting it so that it matches a predefined target.

(From the iZotope web site.)


Isn't that exactly what the Resolve feature is doing? If not, what would a true Normalization feature do differently than what is being done now?

The normalize or clip gain functions would be used on a clip or tracks before any other processing, to either balance one clip against another, and/or bring them to a working level with regard to the unity gain level of the track. Once set you would not change them. Adding FX in the chain such as a compressor or EQ will have an affect on levels which is why plugins have output level controls.
The finished soundtrack can, if needs be, be given a final normalization to bring it to the desired audio standard.
Of course if you go back and change the initial gain of the clip it will affect the fx chain that follows. Here you can see where clip gain has been applied to balance clips against each other.None of this is unconventional.

What is audio normalization?

Normalisation was often regarded as a destructive process, increasing the source's highest peak level to meet a predetermined maximum level - usually assumed to be full scale digital (i.e. 0dBFS and later 0dBTP due to the inherent intersample peaks resulting from the process), hence the term Peak Normalisation. Often Peak Normalisation is simply referred to as Normalisation.

Loudness Normalisation can refer to either perceived loudness (e.g. integrated full programme measurement) or RMS where the source loudness may be either above or below the target level, and where a level correction is required to conform to, for example, a single target (e.g. -23LUFS).

In some implementations, Loudness Normalisation is a destructive process where multiple targets are defined including upper and lower limits for LRA (e.g. for dialogue), with limiters available for True Peak, as well as Momentary and Short Term Maximum, and sometimes this can result in sonic discolouration of the source in the resulting deliverable.

During the QC process, multiple conformance checks are completed including one for Loudness conformance of audio essence in the deliverable and sources are often processed (e.g. a batch process) to conform to the specification outlined in the QC item in the Template used.

Using terminology such as Loudness Conformance in delivery specs, bespoke software tools, and in communication through the media value chain provides improved clarity, rather than using general terminology that may have traditionally been misused or used without reference.

Normalisation was often regarded as a destructive process

Isn't that simply because DAW's are traditionally destructive editors, overwriting the original audio file every time you save?

NLE's don't work that way, they're non-destructive by nature.

But whether a metadata change or a baked-in change, I'm not understanding how the processing is any different. Why would non-destructive Normalization not still be...Normalization? If ProTools was a fully non-destructive DAW, would the A/85 Normalization there be any different than what Fairlight does?

Isn't that simply because DAW's are traditionally destructive editors, overwriting the original audio file every time you save?

Professional post-production audio workstations (such as Sonic Solutions systems - derived from the SoundDroid system which was developed prior to EditDroid) were always non-destructive and featured real-time Timelines, ones that, for example, didn’t require rendering new files when creating fades and crossfades (reduce files were instead continuously updated). Many DSP processes (e.g. NoNoise Manual DeClick) retained a change list rather than processing the source destructively.

Consumer two track audio editors and the subsequent more affordable audio workstations, were not able to provide real-time Timelines and most DSP had to be “baked in” (e.g. AudioSuite processing).

But whether a metadata change or a baked-in change, I'm not understanding how the processing is any different.

Peak Normalisation is usually a process at the source “input” pre-DSP. Fairlight will process the source in the same manner when the Normalise Audio Levels function is executed prior to any additional DSP and therefore functions as a non-destructive Peak or Loudness Normalisation tool.

Fairlight also processes the source at the source's “output” to ensure that regardless of any DSP the source will always conform to the set target. Here the Normalise Audio Levels function acts as a Loudness Conformance tool which is much more practical and useful in current workflows.

Peak Normalisation is usually a process at the source “input” pre-DSP. Fairlight will process the source in the same manner when the Normalise Audio Levels function is executed prior to any additional DSP and therefore functions as a non-destructive Peak or Loudness Normalisation tool.

Fairlight also processes the source at the source's “output” to ensure that regardless of any DSP the source will always conform to the set target. Here the Normalise Audio Levels function acts as a Loudness Conformance tool which is much more practical and useful in current workflows.

Taking the output of the Effects chain and applying the "normalization" algorithm adjusting the input BEFORE the effects chain does not accomplish what you are saying here. It results in the error that I am speaking of that can occur in many use cases, although a few folks insist this highly specific implementation is a "feature". I beg to differ on the "useful feature" part. THAT "useful feature" is a highly specific use case where this works, and I am certain in that use case it is nice. A touch more complexity in the implementation but not necessarily more difficult for the end user would have satisfied ALL use cases that I can think of for this part of things.

My intent was not to be argumentative.

I finally found audio path diagrams but information is lacking.

I regret deeply that I've posted anything.

A single use case is accommodated and that is that.