APU Dynamics Optimizer

APU Dynamics Optimizer is a standalone loudness optimization tool supporting a variety of loudness types, distribution modes, and real-time dynamics controls. It performs an up-front analysis of an entire audio file, building a loudness distribution which can then be reshaped to a target distribution using uniform, normal, source-match, or reference-match modes.

The optimizer takes the source and target distributions and makes many small adjustments to the compress/expand ratio over time, guided by the full dynamic landscape of the file. Traditional compressor controls like attack and release are available and can be adjusted in real-time during playback, along with the target range and distribution shape. Export is to 32-bit float wav at the original sample rate.

• Distribution modes: Uniform, Normal (with skew), Source-match, and Reference-match.
• Loudness types: LUFS (Momentary, Short-term, Integrated), RMS, Peak, and True Peak.
• Multi-channel: Supports stereo, surround, and Dolby Atmos® formats.
• Weighting curves: K-weighting, ECMA-418, ITU-R 468, A-weighting, C-weighting.
• Built-in limiter: Configurable look-ahead with Peak or True Peak mode.

This is a standalone tool. It requires an up-front analysis pass over the entire file, which is not possible in a real-time plug-in context.

Note: Existing customers can request a discount code using the support beacon on the bottom-right.
System requirements: macOS 10.14 (x64, ARM), Windows 10 (x86, x64), OpenGL 3.2.
Supported software formats: Standalone application.

Walkthrough

When you first open an audio file, the optimizer will analyze the audio file’s loudness timeline. You’ll then be presented with a pair of histograms. On the left, you’ll see the audio file’s current loudness distribution. On the right, you’ll see a target distribution. The optimizer’s job is to dynamically compress and expand the audio throughout the entire audio file strategically to match the target distribution.

Before continuing to the next step, you can make adjustments to the source percentiles. This allows you to exclude any unwanted audio from the optimization process. By default, the percentiles are set to match the LRA standard from LUFS. This generally does a good job capturing the main body of the audio.

Once you’re happy with the source percentile range, you can then press the right arrow to continue to the next step.

Once optimization is complete, you can play the result in real-time while adjusting the character of the sound with the same collection of parameters from the compressor plug-in. Use the play/stop buttons and seek slider to navigate the audio while making any adjustments you’d like.

Once you’re happy with the result, you can export the optimized audio to a new file. This file is always stored with the same samplerate as the original audio, using 32-bit floating point wave format. This ensures the highest quality audio is preserved. This format also allows samples over 0 dBFS, which ensures no clipping occurs.

You can listen and export with or without limiting. The built-in limiter can be configured in the Limiter tab, or you can bypass the limiter entirely and use your own external limiter after export.

Dynamics Shaping

The dynamics optimizer allows you to reshape the full dynamics spectrum of your audio file. You can choose between Uniform, Normal, Source, or Reference distribution. The optimizer takes this information and uses it to make many tiny adjustments to the compress/expand ratio over time, achieving a level of redistribution that is not practical with a traditional compressor.

The optimizer can be configured to use the following distribution types:

Uniform

Uniform distribution type is a flat distribution across the entire loudness range. This is a good starting point for most audio files, as it doesn’t favor any particular loudness range. The result is an extremely balanced loudness profile.

Normal

Normal distribution type is a bell curve distribution which can be skewed toward the top or bottom of the distribution. This allows you to dial in a very particular shape of your choosing, whether it’s to more closely match the original audio or for creative effect.

Source

Source distribution type preserves the analyzed source histogram shape while mapping into your configured target range. This is useful when you want to compress or expand dynamics without reshaping the underlying histogram profile.

Reference

Reference distribution type matches the histogram shape from a second analyzed file. Load a reference file after source analysis, then optimize. When the reference file is loaded, target range is initialized to the equivalent reference range and re-synced as source percentiles change.

Measurement Modes

One of the key features of this optimizer is its support for a variety of loudness types. The optimizer’s detection circuitry can be configured to use either LUFS or traditional measurement units (RMS, Peak). You can also choose to split or link channels, which is particularly useful for multi-channel audio formats such as Dolby Atmos®.

The optimizer’s detection circuitry can be configured to use any of the following loudness types:

Momentary

Momentary is a short duration LUFS measurement (400ms), which is useful for processing the audio’s moment-by-moment dynamics. This mode is responsive but also transparent, making it an appropriate default for the optimizer. K-weighting and channel weighting are applied to the optimizer’s detection circuitry, improving perceptual accuracy.

Short-Term

Short-Term is a longer duration LUFS measurement (3 seconds), which is useful for processing the audio’s large scale dynamics. Given the longer duration, this mode is less responsive than momentary, but it can be useful for adjusting the bigger picture of the audio’s dynamics. K-weighting and channel weighting are applied as with momentary.

Integrated

Integrated is an infinite duration LUFS measurement. K-weighting and channel weighting are applied as with momentary, as well as both relative and absolute gating. This loudness type can be useful for continuously wrangling the audio’s integrated loudness toward a specific target range.

RMS

RMS loudness types use a rolling time interval of 150ms by default, which can be adjusted. RMS is a traditional loudness measurement, which is a good first-order approximation of perceived loudness. This mode doesn’t apply K-weighting or channel weighting to the optimizer’s detection circuitry, giving it a more traditional character.

True Peak

True Peak loudness types are similar to traditional peak measurements, but they are more accurate as they take into consideration inter-sample peaks. These measurements use a time interval equal to the configured block size (1ms by default). This mode is extremely responsive, making it suitable for very fast attack and release times.

Peak

Peak loudness types are a standard peak circuit, operating on a time interval equal to the configured block size (1ms by default). This is another traditional compressor style measurement. Technically, this mode is even slightly more responsive than True Peak, but it is less accurate as it doesn’t take into consideration inter-sample peaks.

Limiter

The optimizer includes a built-in limiter which works seamlessly with the optimizer’s attack/release ballistics. The limiter’s look-ahead setting is used to throttle the attack/release ballistics to prevent overshoots and can be configured to operate in Peak or True Peak mode.

Alternate Weighting

In addition to the standard K-weighting for LUFS measurements, the dynamics engine now supports several alternate weighting curves. These curves modify the detector’s frequency response, allowing the dynamics processing to be tailored for specific perceptual goals or technical standards. This can be useful for achieving more transparent results, targeting specific frequency ranges, or emulating the behavior of classic hardware.

The screenshots below are courtesy of the Loudness Contour plugin, which allows visualization of the frequency response of each contour type. For the optimizer, the supported weighting types are implemented as zero-latency IIR filters, which allows them to be used in real-time without introducing latency.

ECMA-418

ECMA-418 is a modern standard for perceptual frequency weighting, designed to reflect human hearing sensitivity in typical listening environments. Unlike legacy curves such as A- or C-weighting, ECMA-418 is based on contemporary psychoacoustic research. Using it in the detector can lead to dynamics processing that aligns very closely with perceived loudness, resulting in transparent and natural-sounding optimization.

ITU-R 468

ITU-R 468 is a weighting standard designed for measuring noise in broadcast systems. It strongly emphasizes the 1-9 kHz range, making the detector highly sensitive to frequencies associated with harshness or sibilance. This can be creatively used as a frequency-selective optimizer to tame harshness in cymbals, vocals, or distorted guitars without needing a separate EQ in the sidechain.

A-weighting

A-weighting approximates human hearing at moderate levels (~40 phon). It has a significant low-frequency roll-off and a broad peak in the mid-range (1-5 kHz). When used in the detector, it causes the optimizer to be less reactive to low-frequency content and more sensitive to the mid-range, which can be useful for managing vocal presence or preventing bass frequencies from dominating the optimization.

C-weighting

C-weighting has a much flatter response than A-weighting, with less aggressive low-frequency roll-off, and is intended for higher sound pressure levels (~100 phon). Using C-weighting results in dynamics processing that is more responsive to low-frequency energy, making it suitable for full-range material where you want the bass to contribute more significantly to the overall dynamics control.

Credits

This software was developed by APU Software, LLC and is available as a standalone application (windows x64/x86, macOS universal). The software libraries below are utilized for portions of the software:

• JUCE (cross-platform audio and user interface framework)
• Boost (header-only algorithms)
• libebur128 (loudness measurements)
• melatonin_blur (blur effects)
• r8brain (sample rate conversion)

Demo video song credits:

• Dan Phillipson - Feel The Drama, licensed via PremiumBeat.