MasVis for Windows

See what loudness, limiting, headroom and dynamics are actually doing.

MasVis for Windows is an independent Windows-native fork of MasVisGtk by ITProjects. MasVisGtk builds on PyMasVis by Joakim Fors, a Python reimplementation of the original MasVis.

It keeps the proven MasVis/PyMasVis analysis and report concepts while replacing the GTK/libadwaita desktop interface with a native PySide6/Qt application for Windows. On top of the classic report, MasVis for Windows adds explainable tools for interpreting one master and comparing two versions of the same music.

We don't replace DR. We explain it.

• Native Windows interface with multiple report tabs, drag & drop and Advanced Open.
• Detailed MasVis reports with waveform, spectrum, crest factor, histogram, Peak-vs-RMS, EBU R128 loudness and TT-style DR information.
• Explainable Dynamics Assessment for single tracks.
• Same-content Dynamics Comparison with reliable-alignment gating, loudness-level matching and separate loudness/peak interpretation.
• Play button that sends the current original file to the default audio player configured in Windows. MasVis itself remains analysis-only.
• Side-by-side visual comparison with synchronized zoom and animated GIF export.
• Light/Dark application theme plus Light/Dark report presentation.
• Save and Save All with per-export format and resolution selection.
• High-resolution raster export to PNG, JPEG, WebP and TIFF, plus raster-backed SVG, PDF and EPS containers.
• Remembered open/export locations for a more natural Windows workflow.
• Built-in Help and Glossary covering both the classic report and the newer dynamics features in end-user language.
• FFmpeg/ffprobe decoding using a fixed, verified FFmpeg 9.0.1 Essentials build bundled with the portable Windows release.
• Bounded-memory True Peak processing that preserves validated results while greatly reducing temporary memory use on eligible input.

Two releases of the same song can sound and measure very differently. One may be quieter with more headroom, another may be pushed louder with compression or limiting, and a vinyl or captured version can show different peaks even when its underlying musical dynamics are very similar.

This is part of why discussions around the Loudness War became so focused on DR values: they are useful, but one number cannot describe every kind of dynamic change.

MasVis for Windows helps separate those ideas instead of turning them into one "good" or "bad" score:

• How loud is the track overall?
• How much peak/headroom-based dynamic range does it measure?
• Does the waveform show signs commonly associated with strong level maximization or limiting?
• If two masters measure differently, is the difference mainly level, loudness dynamics over time, peak/crest structure, or a combination?

The goal is not to tell you which master you should prefer. It is to make the measurements easier to understand.

Analyze a track and get the familiar detailed MasVis report: waveform, spectrum, crest factor, peak-vs-RMS behavior, EBU R128 loudness, TT-style DR, True Peak, LRA, PLR and more.

DR remains visible exactly because it is useful. MasVis for Windows simply adds more context around it instead of treating DR as a universal quality score.

Simple question: How strongly does this waveform show measurable signs associated with level maximization and strong limiting?

Dynamics Assessment combines several already-measured indicators into an explainable Level Maximization Evidence score from 0 to 100 and shows which measurements contributed to it.

The result is deterministic and transparent, but it is not a probability that a track was mastered in a particular way and not a sound-quality grade. TT-style DR and LRA remain visible as context and do not directly add score points.

For the frozen v1 methodology, see docs/DYNAMICS-ASSESSMENT-V1.md.

Simple question: If two versions of the same music measure differently, what is actually different — mainly their level, their loudness dynamics over time, their peak/crest structure, or some combination of those?

Dynamics Comparison aligns two versions of substantially the same material, level-matches them, and then keeps several different measurements separate:

• Measured DR — the normal MasVis/TT-style DR result for each version.
• Loudness Dynamics Similarity (LDS) — a 0-100 point score describing how similarly the aligned Short-Term loudness develops over time.
• Loudness Curve Similarity — the direct Pearson correlation of those aligned Short-Term loudness curves, shown as a percentage.
• Loudness Dynamics Advantage — whether one version shows a wider aligned loudness-dynamics span.
• Peak Structure Difference — a separate 0-100 point description of how strongly peak/crest structure differs after alignment and level matching.

For the frozen v1 methodology, see docs/DYNAMICS-COMPARISON-V1.md.

MasVis for Windows includes three controlled demo files built from the same track. You can listen to them, analyze them separately, and reproduce very different Dynamics Comparison cases yourself.

• 🎧 Eternal Desert — Reference
• 🔥 Eternal Desert — Loudness War
• 💿 Eternal Desert — Vinyl-like transfer simulation

(Eternal Desert is an AI-assisted song created with Suno from lyrics and creative prompting by adventureFAN. The Loudness War version is deliberately louder and more heavily limited. The third file starts from that already compressed Loudness War master and applies a simplified vinyl-like transfer chain that changes peak/crest structure without restoring the lost loudness dynamics.)

That gives the project three complementary demonstrations:

• Reference vs Loudness War: DR 10.4 → 6.5, with a real reduction in aligned loudness dynamics.
• Loudness War vs Vinyl-like transfer: DR 6.5 → 10.4, while Loudness Dynamics Similarity remains 99.3 / 100, Loudness Curve Similarity is 99.77%, and no Loudness Dynamics Advantage is detected.
• Reference vs Vinyl-like transfer: both versions measure DR 10.4, yet Loudness Dynamics Similarity is only 62.9 / 100 and Version A shows a Strong Loudness Dynamics Advantage.

Together, these controlled cases show why measured DR and aligned loudness dynamics should be interpreted as related but distinct properties of the signal. The example audio is separately cleared demonstration media and is not covered by the project's GPL-3.0-or-later source-code license. See examples/README.md for the exact purpose, processing and licensing notes.

🎧 Listen / download: Reference MP3 · Loudness War MP3

The Loudness War version was deliberately made much louder and denser from the Reference as a controlled level-maximization stress test. MasVis for Windows reports DR 10.4 vs 6.5, with the Loudness War version about 6.70 dB louder after alignment and a clear reduction in aligned Short-Term loudness dynamics.

This is the familiar case where a much lower DR value coincides with a real change in the song's measured loudness dynamics.

💿 Listen / download: Loudness War MP3 · Vinyl-like transfer simulation

For the second experiment, the already compressed DR 6.5 Loudness War master was processed through a deliberately simplified vinyl-like transfer chain. No expansion or dynamic restoration was applied. Nevertheless, its measured DR rises to 10.4 — matching the 10.4 of the original Reference.

MasVis for Windows simultaneously reports:

• Loudness Dynamics Similarity: 99.3 / 100 — Extremely High;
• Loudness Curve Similarity: 99.77%;
• Loudness Dynamics Advantage: None detected;
• Peak Structure Difference: 71.4 / 100 — Strong;
• EBU LRA change: effectively 0.00 LU;
• Alignment: Reliable.

The DR meter is measuring a real change in peak/RMS or crest structure, but the higher DR value does not mean that the previously lost loudness dynamics have been restored. A higher DR value from a vinyl or needledrop source therefore cannot, by itself, prove that a more dynamic master was used.

The file name uses “Vinyl Remaster” as a short demo label. This is not an actual vinyl release or a complete physical vinyl-cutting/playback model; it is a controlled signal-processing simulation for demonstrating the measurement problem.

🎧 Listen / download: Reference MP3 · Vinyl-like transfer simulation

The third experiment compares the original Reference directly with that same vinyl-like transfer. Both versions measure DR 10.4, but their aligned Short-Term loudness dynamics are clearly different.

MasVis for Windows reports:

• Measured DR: 10.4 vs 10.4 — essentially equal;
• Loudness Dynamics Similarity: 62.9 / 100 — Moderate;
• Loudness Curve Similarity: 95.51%;
• Loudness Dynamics Advantage: Strong — Version A;
• Level Difference: Version B is 0.86 dB louder;
• Peak Structure Difference: 56.7 / 100 — Moderate;
• EBU LRA change: -3.95 LU in Version B;
• Alignment: Reliable.

This is the complementary case: essentially equal measured DR values do not necessarily imply equal loudness dynamics. Here the Reference retains a substantially wider aligned Short-Term loudness span even though both versions receive the same TT-style DR value.

All three experiments are practical validation examples, not proof that the program can reconstruct mastering history. MasVis for Windows reports what the measured signals support and keeps uncertainty visible when they do not support a reliable conclusion.

MasVis for Windows is an analysis tool, not an automatic mastering judge.

It does not:

• rate whether a song sounds good or bad;
• prove which release came first or reconstruct mastering ancestry;
• treat a higher DR value as automatically better;
• treat vinyl/needledrop DR as proof of greater musical dynamics;
• use its 0-100 dynamics scores as statistical probabilities;
• contain its own audio player.

Use the measurements as evidence and context — then use your ears for preference.

The Windows interface currently accepts these extensions:

WAV, FLAC, MP3, M4A, OGG, OPUS, AAC, AC3, AIFF, AIF, AMR, ALAC, PCM, WMA, APE

Actual decoding is performed by FFmpeg. A file still has to contain a stream that the validated FFmpeg build can decode.

The public Windows release is available as a normal Setup installer and as a portable 64-bit ZIP. Both contain the same validated MasVis for Windows runtime and the exact bundled FFmpeg/ffprobe build.

1. Download MasVis-for-Windows-1.1.1-Setup.exe from the Releases page.
2. Run the installer.
3. Start MasVis for Windows from the Start menu or the optional desktop shortcut.

The installer places the application under Program Files, creates a normal Windows uninstall entry and can update an existing MasVis for Windows installation in place in future releases. It does not register MasVis as the default application for audio files.

The current installer is not code-signed, so Windows may identify the publisher as unknown when the downloaded Setup is launched.

Prefer not to install anything? Download MasVis-for-Windows-1.1.1-win64.zip, extract it to a normal folder and run MasVis-for-Windows.exe.

Python, PySide6/Qt, NumPy, SciPy, Matplotlib and the other tested runtime dependencies are included in both distributions. The exact validated Gyan FFmpeg 9.0.1 Essentials ffmpeg.exe and ffprobe.exe binaries are bundled as well, so no separate FFmpeg installation or PATH configuration is required.

The tested release target is Windows 11 x64.

• Windows 11 x64 is the current development/test baseline.
• Python 3.11 or newer; current pinned Windows packaging baseline: Python 3.14.7.
• FFmpeg and ffprobe on PATH, or verified local copies under vendor/ffmpeg/.

Python dependencies are pinned in requirements.txt. Source runs prefer verified binaries under vendor/ffmpeg/ when present and otherwise fall back to PATH. Packaged builds use only the verified FFmpeg binaries bundled with the application and deliberately do not fall back to an arbitrary system FFmpeg.

python -m venv .venv
Set-ExecutionPolicy -Scope Process -ExecutionPolicy Bypass
.\.venv\Scripts\Activate.ps1
python -m pip install -r .\requirements.txt
python .\app.py

MasVis for Windows would not exist without the work that came before it:

MasVis -> PyMasVis -> MasVisGtk -> MasVis for Windows

• MasVisGtk — ITProjects — direct upstream project and analysis/UI basis.
• PyMasVis — Joakim Fors — Python reimplementation of the original MasVis.
• MasVis — original project/research lineage.

This repository is an independent derivative/fork of MasVisGtk and links back to the upstream project explicitly. MasVis for Windows is developed independently and is not an official MasVisGtk release.

Project direction, feature design, testing, and release decisions: adventureFAN.
Most implementation, debugging, and technical documentation: developed collaboratively with ChatGPT.

MasVis for Windows is distributed under the GNU General Public License, version 3 or (at your option) any later version (GPL-3.0-or-later). See LICENSE.

Reused or modified upstream source retains its original copyright notices. The Windows fork's first-party modifications are Copyright (C) 2026 adventureFAN.

Packaged releases also contain third-party runtime components under their own licenses, including Qt for Python/PySide6 and the validated GPLv3 Gyan FFmpeg 9.0.1 Essentials build. See THIRD_PARTY_NOTICES.md for attribution, exact build identity and source-access information.

• Project: https://github.com/adventureFAN/MasVis-for-Windows
• Upstream: https://github.com/itprojects/MasVisGtk