A needle VU meter for Linux and Windows. VST3, CLAP, LV2 and standalone, from a single codebase. Built with DPF and NanoVG.
Leia em português.
My name is Ricardo Rossati, better known as Mister RickRoss. I have been a bass player for 29 years, I live in Brazil. I produce, I record, I used to teach bass at School of Rock, I have played in bands, backing famous national artists. In my career I have had experience with radio, television, stage, touring, recording.
And even having reached that level, here in Brazil I feel music is a path for few, and that musicians do not get the space they deserve. I used to use Studio One on Windows, and it was always all paid. At the music school where I taught, the owner decided to install Linux. My history with technology goes back to Windows 3.11, where I used to install Doom.
Windows never offered me the practicality that Linux offered. And when I got to Linux, I missed several native plugins, and I did not want to keep installing or cracking plugins. With the knowledge I built up through AI, I started coding, it has been 3, 4 years of coding now, and now with Claude some things got a lot easier.
Besides ROSS VU I have Substance, in continuous production progress, I have an app called BandMate, and I was the first Brazilian to release a music video made with artificial intelligence. You can check all of that at misterrickross.com.
The idea for this plugin came, and yes, Claude helped me code it. The needle model and the tests that prove the numbers are mine, Claude helped me with the code around them. All the prototyping was done with a lot of care, the calculations are all on the site. I am inviting everyone to use the plugin and also to contribute to this new model, this new era, where each person can contribute from home, without spending too much time and energy.
Mister RickRoss / Ricardo Rossati misterrickross.com
This is not a decorative pointer animated by RMS. The movement is a real second order system with average rectification, exactly like the electromechanical instrument:
| measurement | VU standard | ROSS VU (measured) |
|---|---|---|
| time to 99% deflection | 300 ms | 299 to 302 ms |
| overshoot | 1% to 1.5% | 0.86% to 1.00% |
| sine at reference level | 0.00 VU | 0.000 VU |
| tone 6 dB below reference | -6.00 VU | -6.012 VU |
Constants: zeta = 0.826, omega0 = 14 rad/s. Verified at 44.1 kHz, 48 kHz and 96 kHz.
The dial scale is drawn from the linear deflection of the movement,
p = 10^(VU/20) / 10^(3/20), which is what produces the compression on the left
and the spread near zero. It is not a scale invented to look vintage.
Because it is average rectified and not RMS, it reads high crest factor material below RMS, just like an iron vane VU. That is correct behaviour, not a bug.
The lamp and the readout are dBTP, via 4x polyphase oversampling (12 coefficients per phase, Blackman-Harris windowed sinc, in the spirit of ITU-R BS.1770-4). The classic case, full scale sine at fs/4 with 45 degree phase:
| reading | |
|---|---|
| sample peak (what almost every meter shows) | -3.01 dBFS |
| ROSS VU | -0.20 dBTP |
| true value | 0.00 dB |
Error within the 4x tolerance of BS.1770. A sample peak meter would be 3 dB wrong on that signal, and that is exactly where the limiter blows up.
-18 dBFS = 0 VU is a tracking and mixing convention, inherited from analog
0 VU = +4 dBu. A finished master does not live there: it sits 4 to 6 dB above,
so it hits the red and pins the needle. That is the right instrument telling the
truth, not a fault.
| what you are looking at | calibration |
|---|---|
| tracking, input gain | -20 or -18 |
| mixing, buses and channels | -18 or -16 |
| finished mix before mastering | -14 |
| commercial master, A/B reference | -12, -10 or -8 |
The same -14 dBFS RMS master reads +4.0 VU at CAL -18, 0.0 VU at CAL -14 and
-4.0 VU at CAL -10. Measured, not estimated.
In REAPER the FX chain is pre fader. If you pulled down the fader on your reference track to level match an A/B, the meter still sees the full level, not what your ears are hearing.
A lamp that only lights after you already clipped is useless. Its job is to show the transient the slow needle hides, so it has to fire before the clip. The threshold is chosen by clicking, in dBTP:
-12 · -9 · -6 · -3 · -1 · -0.1
Default is -6 dBTP. Hold time is deliberately short, 400 ms, so it reads as a blink and not as a stuck lamp. Measured: a 30 ms overshoot above threshold lights it for 430 ms, six overshoots give six blinks, and nothing below the threshold lights it at all.
| control | range | note |
|---|---|---|
| CAL | -20 / -18 / -16 / -14 / -12 / -10 / -8 dBFS | click cycles; defines what 0 VU is |
| RESPONSE | 600 ms to 100 ms | centre gives the normative 300 ms |
| PEAK | -12 / -9 / -6 / -3 / -1 / -0.1 dBTP | click cycles; blinks for 430 ms |
| INPUT | -24 to +24 dB | changes the audio, not just the meter |
| OUTPUT | -24 to +24 dB | output gain |
| MODE | L+R / LEFT / RIGHT / SIDE / L | R | click the label at the bottom left of the dial |
In L | R mode a second pointer appears, thinner and lighter, carrying the right channel. The left one stays the solid pointer.
Presets and mode live in the interface itself, not buried in the host parameter
list. The ‹ MIX BUS -18 › selector in the header steps through the list, and
touching any control switches it to CUSTOM. Everything clickable lights up on
hover, so you can find it without a manual.
| preset | 0 VU | response | mode | lamp |
|---|---|---|---|---|
| MIX BUS -18 | -18 dBFS | 300 ms | L+R | -6 dBTP |
| TRACKING -20 | -20 dBFS | 100 ms | L+R | -12 dBTP |
| MIX READY -14 | -14 dBFS | 300 ms | L+R | -3 dBTP |
| MASTER -10 | -10 dBFS | 300 ms | L+R | -1 dBTP |
| REFERENCE -12 | -12 dBFS | 300 ms | L+R | -1 dBTP |
| STEREO L | R | -18 dBFS | 300 ms | L | R | -6 dBTP |
| PROGRAM SLOW | -18 dBFS | 600 ms | L+R | -6 dBTP |
Prebuilt binaries are on the releases page.
tar xf ROSS-VU-1.4.1-linux-x86_64.tar.gz
cd ROSS-VU-1.4.1
mkdir -p ~/.vst3 ~/.clap ~/.lv2
cp -r ROSSVU.vst3 ~/.vst3/
cp ROSSVU.clap ~/.clap/
cp -r ROSSVU.lv2 ~/.lv2/
Needs glibc 2.27 or newer: Ubuntu 18.04, Debian 10, Fedora 28 and anything
after. Version 1.4 needed glibc 2.43 and failed to load on most distros with
GLIBC_2.43 not found. 1.4.1 fixes that and changes nothing in the metering.
In REAPER: Options > Preferences > Plug-ins > VST > Re-scan. CLAP in ~/.clap
is scanned along with it.
Unzip and drop ROSSVU.vst3 and ROSSVU.clap into
C:\Program Files\Common Files\VST3 and ...\CLAP.
ROSS VU builds against DPF, which sits next to it as a sibling folder. DPF
carries its own submodule (pugl), so it is the DPF clone that needs
--recursive, not this repo (this repo has no submodules).
git clone https://github.com/RickRossati/ross-vu
git clone --recursive https://github.com/DISTRHO/DPF
cd ross-vu
make # VST3 + CLAP + LV2 + standalone, into ../bin
./instalar.sh # or this: builds, then copies into ~/.vst3, ~/.clap, ~/.lv2
Known-good DPF commit: 4238e1c (Sep 2026). DPF's develop branch moves; if a
later commit breaks the build, check that one out inside DPF.
A plain make links against the glibc of the machine you build on, so the
binary only runs there or on something newer. Release builds for Linux use
./build-linux-compat.sh, which compiles inside Ubuntu 20.04 with podman and
generates the LV2 .ttl files (a separate step in DPF, easy to forget, and
without them no host finds the LV2). Output in ../bin-compat.
make windows cross compiles with mingw-w64 and produces VST3 + CLAP in
../bin-win. The binary is PE32+ x64 and depends on no mingw DLL, only on
Windows' own.
- macOS build (the code is portable, the toolchain is missing)
- Validation of the Windows build in a real DAW
- Level history over time
- Fixed step resizing (today it is free, keeping the aspect ratio)
Honest caveat: neither the Windows .exe nor the .zip has been opened in a
DAW on real Windows, because the machine this is built on does not run Windows.
It was verified as a valid PE with the right exports (GetPluginFactory for
VST3, clap_entry for CLAP). Testing under Wine failed before the plugin even
ran: Wine's wglChoosePixelFormatARB returns zero formats and pugl gives up
there. That hits any DPF app under Wine on this machine, it is not a ROSS VU bug.
If you run it on Windows, telling me what happened is genuinely useful.
Not talk. The tests live in tests/, and tests/rodar.sh runs all three:
vutest2.cppextracts the actualstruct MovementfromRossVUPlugin.cppand measures rise time, overshoot and rest level at 44.1k, 48k and 96k.tptest.cppextracts the realstruct TruePeakand compares it against the analytic peak on six signals, including the fs/4 case.lamptest.cppextracts the realstruct PeakLampand counts blinks and duration across six combinations of overshoot and threshold.- End to end: a 1 kHz tone generated at -18.01 dBFS RMS, routed straight into the plugin's input ports through PipeWire, needle settling at 0.0 VU and true peak at -15.0 dBTP.
GPL-3.0-or-later. DPF is ISC. Liberation Sans is OFL.
Built by Mister RickRoss, 2026.