Automatic Level Normalization for All Neural Captures

Quad Cortex Feature Requests
1

Automatic Level Normalization for All Neural Captures

I’d love to see a standardized reference output level built into the Neural Capture process for all captures, including third-party and user-created captures.

Currently, different captures can load at dramatically different volumes. This creates a few problems:

  • A/B comparisons become less reliable because louder captures can naturally sound fuller, more exciting, or simply “better.”
  • Auditioning captures within an existing signal chain becomes cumbersome. Some captures need significant boosting just to reach the expected level, while others need to be turned down to prevent clipping downstream blocks.
  • Quickly comparing captures from different creators becomes much harder than it should be because level has to be adjusted manually before you can really evaluate tone and feel.

It seems like the Neural Capture process itself would be the ideal place to address this.

Since Quad Cortex already sends a known test signal through the amp during capture, could it also measure the resulting output and calculate a standardized reference-level offset?

For example:

Capture → validation → reference-level measurement → automatic output trim → save

The actual capture/model wouldn’t need to be altered. QC could simply store a recommended output offset as metadata with the capture. That value would remain attached to the capture when it is shared, sold by a third-party creator, or uploaded to Cortex Cloud.

That would mean a capture made by Neural DSP, a commercial third-party capture creator, or an individual QC user could all arrive at a reasonably consistent and predictable nominal starting level.

I’d favor one universal reference level, rather than separate recording and live standards. Preset levels and global outputs could still be adjusted appropriately for recording, FOH, power amps, etc. The purpose here would simply be to give every capture a consistent starting point.

Ideally, normalization could also be user-selectable with something like “Normalized Capture Level: On/Off.” It could be particularly useful when auditioning captures from Cortex Cloud, where large level differences make rapid comparisons difficult.

This wouldn’t need to make clean, crunch and high-gain amps equally loud or alter their natural dynamics. The goal would be to establish a consistent nominal reference so we can more easily hear the differences between captures and choose based on tone and feel rather than level.

In other words: standardize the starting point, not the sound.

I think this would make capture comparisons much more meaningful, dramatically improve Cortex Cloud auditioning, reduce unnecessary gain-staging adjustments, and make swapping captures within existing signal chains much easier.

2

Hello and welcome. Clearly you’ve put a lot of thought into this thread, but this request already exists:

3

Thanks — I saw that thread, but I think what I’m proposing is importantly different from post-capture normalization.

I agree that taking an already-created quiet Capture and simply boosting its output afterward can raise the noise floor. My suggestion is to establish a standardized target output level during the Capture process itself, while the QC is calibrating and analyzing the source.

In other words, rather than capturing at an arbitrary level and normalizing afterward, the Capture process would determine the appropriate gain staging necessary to create the Capture at the target level in the first place. That should avoid the noise-floor problem you’re describing.

Ideally, Neural could define that target as part of the Capture standard so that both Neural and third-party Captures land in a reasonably consistent output range.

The goal isn’t really “normalization” in the traditional sense — it’s standardized gain staging at the point of Capture.