Tools and Recommendations

The best free LUFS meters for voice (and when REAPER's own is enough)

By StudioRescue Editorial

Published October 4, 2026

“Which LUFS meter should I use?” is usually the second question. The first one, whether the episode needs to be at a particular loudness at all, has a short answer: podcast hosts and streaming platforms normalize what you send them, so the number matters mostly for consistency between your own episodes. Apple, for example, asks for overall loudness around -16 dB LKFS with about 1 dB of tolerance and a true peak that does not exceed -1 dB FS, calculated to the ITU-R BS.1770-5 recommendation.

Once you accept that there is a number to hit, you need something that reads it. This article defines what to look for, says plainly when REAPER already covers it, and then compares the free meters that are worth the install.

The criteria, set before picking anything

Comparison articles are only useful if the standard is fixed in advance. For a voice or podcast workflow, a meter has to:

• Be genuinely free, with no time limit and no feature you need held behind a paywall.
• Measure to the published standard, which means ITU-R BS.1770 and EBU R 128, not a house approximation.
• Run as a plugin inside REAPER on Windows or macOS, so it sits on the master bus where you work.
• Show integrated loudness plus at least short-term, because the integrated figure alone tells you nothing about where the problem is.
• Leave the audio alone. A meter that also changes level is a different tool with a different risk.

What a loudness meter is actually showing you

Most of the confusion around LUFS comes from reading one number when the meter is showing three. Under EBU R 128 the three loudness readouts are measured over different windows:

The gating on the integrated figure matters for voice more than people expect. R 128 applies an absolute gate that treats anything below -70 LUFS as silence, and a relative gate that discards material falling 10 LU below the running integrated value. Momentary and short-term are not gated. That is why a conversation with long pauses still reports a sensible integrated number: the silence is excluded rather than averaged in.

Two other readouts show up on every meter in this article. Loudness range (LRA) is a statistical description of how much the loudness varies across the programme, and the standard deliberately sets no maximum. True peak is the inter-sample peak, which is the figure delivery specs put a ceiling on.

Start with what REAPER already does

REAPER, currently at version 7.81 released September 28, 2026, has loudness measurement built in, and for a lot of workflows it is the whole answer.

The render dialog can calculate loudness as part of rendering, and it can do that as a dry run, reporting the figures without writing media to disk. Version 7.80 added actions to calculate the stereo loudness of selected tracks via a dry run render, and 7.75 added the equivalent for the source media of selected items. REAPER 7.81 even added the option to render with no loudness calculation at all, which tells you how routine the calculation has become. REAPER also ships its own loudness meter JSFX, covered in section 6.18 of the User Guide, so there is a plugin on the FX list before you download anything.

If your question is “what does this finished file measure, and did I hit the target”, that is a dry run and you are done. The step by step is in our guide to getting a podcast to -16 LUFS in REAPER, which covers the normalization and true-peak settings in the same dialog.

When a free meter is worth installing

Three situations make a third-party meter worth the download:

1. You want it live. A plugin on the master shows momentary and short-term as you play, so you can see the quiet patch while you are listening to it rather than after a render.
2. You are working to a readout you need at a glance. A large, legible meter you can park on a second monitor changes how often you actually check.
3. You need dialogue-gated measurement. This is a genuinely voice-specific feature and only one of the three free options below has it.

Three free meters worth having

Youlean Loudness Meter 2 (free version)

The one most people land on, and a reasonable default. The free version gives you the readouts the job needs: loudness units, integrated, true peak, loudness range, peak to loudness ratio and dynamic range, with a resizable window, a mini view, and mono, stereo and 5.1 support. It runs as VST2, VST3, AU and AAX on Windows and macOS.

What you give up by not paying: the streaming presets, which is the one that stings, since those are what put a target line on the meter for you. Pro also adds a dynamics graph, PDF, PNG and SVG export, saved custom presets, themes, A/B save states and drag and drop file analysis, for $39 plus tax as a one-time purchase. If you only ever work to one target you can remember, the free version is enough. If you deliver to several platforms with different targets, the presets are the thing you will miss.

TBProAudio dpMeter 5

The pick for spoken word, and the most capable of the three on paper. Version 5.3.9, last updated June 2026, free with no paid tier, no nag screen and no registration. It covers RMS (integrated, momentary, peak, true peak and crest), full EBU R 128 (integrated, short-term, momentary, true peak and loudness range), and a dialogue-gated mode that reports dialogue-gated integrated loudness, dialogue proportion and dialogue-gated loudness range. Windows builds are VST, VST3 and AAX; macOS builds are VST, VST3, AU and AAX as universal binaries for Intel and Apple silicon. There are GPU and non-GPU versions, which is worth knowing if your session is already working the CPU hard.

The dialogue-gated readouts are the reason this is the voice recommendation. For a programme that is mostly speech with music under it, a dialogue-gated measurement describes the speech rather than the average of speech and music, which is closer to what a listener is actually judging. It is a less pretty interface than Youlean's. That is the trade.

MeldaProduction MLoudnessAnalyzer

Melda describes it as an EBU R 128 and ITU-R BS.1770-3 compliant loudness meter, with a peak meter, momentary, short-term and integrated loudness, a loudness range meter, and the EBU+9, EBU+18 and EBU+27 scales. It is free, and the practical caveat is how you get it: it comes as part of Melda's free MFreeFXBundle rather than as a single download, so you are installing a bundle to get one meter. If you already run Melda's free plugins, you may have it on your system right now.

A quick way to pick

If a new plugin does not appear in REAPER's FX browser after installing, that is a scan path problem rather than a plugin problem, and it is covered in how to install VST plugins in REAPER.

What a loudness meter will not do

This is the part most roundups leave out, and it is the part that saves you time.

• It will not change your loudness. Every meter here is analysis only. Youlean's own documentation says the plugin cannot alter audio loudness. Reading -19 LUFS does not move you to -16; you still need normalization on render, or a gain move.
• It will not tell you the recording is good. A file can measure exactly on target and still be noisy, harsh or boomy. Loudness is a level measurement, not a quality measurement.
• It will not even out two voices. Integrated loudness is one figure for the whole file, so a quiet guest and a loud host average out to something that looks correct and sounds uneven. That is a leveling job, covered in matching podcast host and guest loudness in REAPER.
• It will not stop a platform renormalizing you. Hitting a target is about consistency and headroom, not about preventing the platform from adjusting playback level. If your upload came back sounding smaller than your mix, see why vocals sound thin and quiet after uploading.
• It is not an audiobook compliance check. Audible's ACX requirements are not written in LUFS, so a file normalized to a podcast loudness target can still fail submission. If Audible is the destination, work from how to meet ACX audiobook specs in REAPER instead.
• A single number will not tell you where the problem is. That is what short-term and LRA are for.

If two meters disagree

All three meters above state compliance with the same measurement standard, and REAPER's own calculation targets the same specification. So the integrated number is not really what separates these tools, and a disagreement is usually a question about what was measured rather than about the meter. Before assuming one is wrong, check the ordinary causes:

1. You measured different spans. An integrated reading covers from when the meter was reset to when you stopped. A render measures the rendered region. If one included four seconds of silence at the top and the other did not, they are answering different questions.
2. You compared integrated with short-term. These are different readouts and they are supposed to differ. Confirm which label you are reading on each.
3. The meter was not reset. An integrated figure accumulates. If you did not reset it before the pass, it still carries the last one.
4. The plugin sat in different places. A meter before a limiter and a meter after it will not agree, and neither will a meter on a track versus one on the master.
5. Mono versus stereo. Channel configuration affects the measurement, so compare like with like.

If you want the figure that matters for delivery, take it from the rendered file rather than from a live meter mid-session. That is the file the platform will measure.