Hit −23 LUFS Without Crushing the Read: Voice‑Over Loudness for Pros

Master voice-over to an integrated loudness of around −23.0 LUFS for broadcast or −20 to −14 LUFS for most streaming platforms, and keep true peak below approximately −1 to −2 dBTP. These are two separate measurements: loudness and peak level both need checking, ideally with a BS.1770 or EBU-compliant meter, followed by a quick critical listen before you export.

---

TL;DR:

• Master voice-over loudness should follow specific platform targets, typically around −23.0 LUFS for European broadcast or between −20 and −14 LUFS for streaming.
• Re-measure true peak after encoding, as lossy codecs can introduce peaks that exceed the maximum true-peak ceiling and cause clipping.
• Achieving compliance requires careful gain staging, gentle compression, and proper gating, especially for long-form content, to preserve vocal clarity and natural dynamics.
• Consistently matching the loudness target for each platform prevents inconsistency across media types like TV, radio, podcasts, and streaming services.

---

Table of Contents

• Standards and targets: EBU R128, ITU BS.1770, ATSC and streaming conventions
• How to measure loudness correctly for voice-over
• Practical mastering chain for voice-over: stepwise processing to hit loudness without ruining dynamics
• Delivery checklist and QC: what to include with final voice-over files
• Publisher perspective and practical checks
• Differences in loudness standards between regions
• Implications of loudness standards on voice-over consistency across different media
• Common challenges and pitfalls when applying loudness standards to voice-over recordings
• Examples of compliant vs. non-compliant voice-over audio
• What voice-over loudness standards actually reward
• Where to find voice talent that delivers to loudness spec
• Sources
• FAQ

Standards and targets: EBU R128, ITU BS.1770, ATSC and streaming conventions

Three documents govern almost every loudness conversation in professional audio. The EBU R128 recommendation sets a program loudness target of −23.0 LUFS for broadcast delivery in Europe, with a measurement tolerance of ±0.2 LU and a wider ±1.0 LU allowance for live programming. ITU-R BS.1770-5 defines the actual measurement algorithm behind that number, including the gating method and the true-peak calculation that every compliant meter runs on. LUFS and LKFS describe the same measurement, just from different standards bodies.

In the United States, ATSC A/85 governs content exchange and recommends default target loudness values along with true-peak ceilings, typically below −2 dBTP, for material delivered without loudness metadata.

Streaming complicates the picture because there is no single global rule.

• Broadcast workflows commonly follow the EBU R128 baseline of −23.0 LUFS.
• Streaming distribution often sits between −20.0 and −16.0 LUFS when metadata is not managing playback gain, according to EBU guidance on streaming loudness.
• Individual platforms have their own typical loudness levels for music and podcast contexts, with around −14 LUFS being a common practical convention rather than a universal specification.

Loudness targets vary by delivery path, and the number that matters is whatever the EBU streaming supplement or the client brief specifies for that platform. Always confirm the actual delivery target before mastering rather than assuming a default.

How to measure loudness correctly for voice-over

Accurate measurement depends on using the right meter and reading it the right way.

1. Choose a meter that implements ITU-R BS.1770-5 and switch it to EBU mode when that option exists, then confirm the gating and channel configuration match your source file.
2. Watch all three measurement windows: momentary (400 milliseconds), short-term (3 seconds) and integrated (start to stop). The short-term window is often the most useful for checking foreground speech level in real time.
3. For long-form spoken content, use dialogue-gated measurement so silences and background elements do not skew the reading; for short-form spots, relative gating against the whole clip is usually sufficient.
4. Measure true peak in dBTP rather than relying on a sample peak meter, since sample peaks miss the intersample peaks that appear after decoding.
5. Re-measure true peak after any lossy encode, because compression codecs can introduce new peaks that were not present in the source file.

Pro Tip: Keep a small allowance, roughly ±0.2 to 2 dB depending on your workflow, for natural measurement variation between meters, and note the exact figures you hit so clients have a record.

Practical mastering chain for voice-over: stepwise processing to hit loudness without ruining dynamics

Voice-over mastering has one job that differs from music mastering: preserve every consonant while still hitting a number. That means the chain should be conservative at every stage rather than relying on one aggressive plugin at the end.

Start with gain staging. Record and edit with peaks comfortably below 0 dBFS so there is real headroom left for true-peak limiting later in the chain. Apply corrective EQ first: a high-pass filter to remove low-frequency rumble, then targeted cuts on resonant frequencies that make a voice sound boxy or nasal. Only add presence or air boosts afterward, and only if the recording genuinely needs it.

Compression comes next, and it should even out delivery without flattening it. A gentle ratio in the 1.5:1 to 3:1 range, a moderate attack that lets the front of words through, and a release tied to the natural pace of the read will usually beat a fast, hard-limiting setting that squashes consonants.

• Use a de-esser or narrow multiband treatment for sibilance rather than broadband limiting.
• Save the true-peak limiter for the final stage, set to the delivery ceiling, commonly −1 to −2 dBTP depending on the spec.
• Measure the voice-only stem against the target LUFS before adding music, sound effects or any other layer.
• Re-check overall program loudness once music and effects are mixed in, since combined elements change the integrated reading.

Pro Tip: Master the dry voice stem to spec first. It gives you a clean reference point and makes it far easier to spot what the music bed or sound design is doing to the final number.

Delivery checklist and QC: what to include with final voice-over files

A finished voice-over file should arrive with enough documentation that a client or downstream engineer never has to guess whether it hits spec.

1. State the measured integrated LUFS and true peak in dBTP directly in the delivery notes or a QC screenshot from your meter.
2. Confirm sample rate, bit depth, codec and whether stems are being delivered alongside the mixed file.
3. Include loudness metadata, such as dialnorm tags, when the delivery spec calls for it; otherwise follow the exchange-without-metadata defaults described in ATSC's modernized A/85 guidance.
4. Re-encode to the final delivery codec and re-measure true peak afterward, since encoding can push peaks past the ceiling even when the pre-encode file was clean.
5. Add a short technical note listing the meter model and mode used, the measurement windows referenced, and a summary of the processing chain applied.

Publisher perspective and practical checks

Meters give a repeatable number, but that number is still an estimate. ITU-R BS.1770-5 itself notes that measured loudness depends on the material and the listening conditions, which is why critical listening in a reference environment stays part of the process rather than an optional extra.

• Meter readings estimate perceptual loudness rather than replacing a trained ear in a treated room.
• A platform works with vetted voice actors and applies a filtering and vetting process so clients can find talent capable of delivering to a defined spec.
• The platform reports a high client satisfaction rate across completed projects, reflecting consistent delivery quality rather than loudness compliance specifically.

Documented QC, paired with a listen on real monitors, catches the gap between what a meter reports and what an audience actually hears.

Differences in loudness standards between regions

The clearest split in global loudness practice runs between American and European broadcast conventions. ATSC A/85 governs content exchange in the United States and recommends default target loudness values for material delivered without metadata, along with true-peak ceilings commonly set below −2 dBTP in that scenario. EBU R128 sets a European baseline of −23.0 LUFS with a tighter ±0.2 LU tolerance for standard measurement.

The practical difference for a voice actor or engineer is less about the exact numbers and more about how metadata is handled. ATSC's guidance leans on metadata, such as dialnorm, to manage loudness through the broadcast chain, while EBU R128 favors normalizing the program itself to the target before delivery. A voice-over mastered correctly for one region's convention will not automatically clear the other region's true-peak ceiling or metadata expectations.

This matters most for voice actors and studios working across borders, dubbing a campaign for multiple territories or delivering the same narration to broadcasters in different countries. The safest approach is treating every regional spec as a distinct delivery target rather than assuming one master file covers every market. When a brief does not specify a region's standard, ask before mastering, since retrofitting true-peak ceilings or metadata after the fact costs more time than building them in from the start.

Implications of loudness standards on voice-over consistency across different media

The same voice-over often travels across TV, radio, podcasts and streaming, and each medium carries its own loudness convention. A commercial mastered to a broadcast target of −23.0 LUFS under EBU R128 will sound noticeably quieter than the same read mastered for a podcast platform working closer to −16 LUFS.

Radio has historically applied its own gain management on the transmission side, which means a file delivered too hot can clip further down the chain even when it measured cleanly in the studio. Podcasts and streaming, by contrast, increasingly rely on platform-side loudness normalization, so a mismatch between your master and the platform's target results in automatic gain changes the voice actor never hears during their own session.

The practical implication is that a single "one true master" rarely survives multi-platform delivery. Voice actors and engineers who repurpose the same read across formats need a per-platform export step, not just a single final bounce. Consistency, in this context, means matching each destination's convention rather than forcing one loudness value to work everywhere. Skipping that step is one of the most common reasons a voice-over sounds perfect in the studio and inconsistent once it reaches an actual listener.

Common challenges and pitfalls when applying loudness standards to voice-over recordings

The most frequent mistake is confusing loudness with peak level. A file can measure correctly at −23.0 LUFS integrated and still clip on playback if true peak was never checked, because BS.1770-5 treats these as two distinct measurements that both need controlling.

Over-compression is a close second. Voice actors and engineers under pressure to hit a loudness number sometimes reach for a hard limiter instead of proper gain staging, and the result is a read that hits spec numerically while losing the articulation that made it worth recording in the first place.

Gating errors also trip up long-form work. Measuring an audiobook chapter or a long narration without dialogue gating can produce a skewed integrated reading, since silences and background noise pull the average away from where the actual foreground speech sits.

Finally, many pitfalls appear only after encoding. A true-peak reading taken before a lossy export can look perfectly safe, then the encoded file introduces new intersample peaks that push past the ceiling. Re-measuring after encoding, not just before, is the step most often skipped under deadline pressure.

Examples of compliant vs. non-compliant voice-over audio

A compliant broadcast voice-over measures at −23.0 LUFS integrated, sits below −1 dBTP true peak, and was checked again after the final encode to confirm nothing shifted. The read itself still has natural dynamic movement: quieter phrases and emphasized words are audible without needing the listener to reach for the volume knob.

A non-compliant version of the same script often measures loud on a sample peak meter but fails a true-peak check, revealing intersample peaks that a basic meter never caught. Another common failure pattern is a file that hits the correct LUFS number but sounds flat and fatiguing because heavy limiting was used to get there quickly, erasing the natural rise and fall of the performance.

The difference rarely comes down to the read itself. It comes down to whether the mastering stage measured both loudness and true peak, gated the measurement appropriately for the content length, and verified the numbers again after the final export. Two voice-overs with identical scripts and similar performances can land on opposite sides of compliance purely based on how carefully that last stage was handled.

What voice-over loudness standards actually reward

The industry conversation around loudness standards spends a lot of time on target numbers and not nearly enough on what those numbers cost a performance if applied carelessly. A voice actor who chases −23.0 LUFS with a hard limiter has technically hit the brief while quietly destroying the thing that made the read worth buying: its dynamic shape.

Conventional advice treats loudness compliance as a mastering afterthought, something to fix in the last five minutes with a limiter plugin. That gets the order backward. Gain staging and restrained compression earlier in the chain do more for both loudness accuracy and vocal clarity than any limiter setting chosen under deadline pressure.

If there is one priority worth adopting before any other, it is measuring the dry voice stem against target before anything else touches the session. Everything downstream, EQ choices, compression ratios, the final limiter ceiling, gets easier to judge once you know exactly where the voice itself sits.

— Onur

Where to find voice talent that delivers to loudness spec

Getting the mastering chain right solves half the problem. The other half is working with a voice actor who understands what "deliver to −23 LUFS with true peak under −2 dBTP" actually means before they hit record. VoiceBros connects clients directly with vetted voice actors across commercials, e-learning, audiobooks, IVR and dubbing, so the technical spec is part of the brief from the start rather than something fixed after the fact.

• Browse vetted voice actors by language, accent and category before placing an order.
• Request delivery-ready files mastered to your specified loudness target, with QC details available on request.
• Compare live demos side by side, then order directly through the platform with pricing and usage rights stated upfront.

Check current pricing and turnaround on VoiceBros before your next project goes out for quote.

Sources

• EBU: Loudness (R128 and related tech docs)
• ITU-R BS.1770-5 (Algorithms to measure audio programme loudness and true-peak audio level)
• ATSC A/85 (Tech RP, 2026 modernization)

FAQ

How loud should voice overs be?

Most broadcast voice-over targets an integrated loudness of −23.0 LUFS under EBU R128, while streaming platforms often expect something between −20 and −14 LUFS. The correct number always depends on the specific delivery brief, so confirm the target before mastering.

Is 10 LUFS too loud?

Yes, a level considerably louder than any standard broadcast or streaming target discussed here, since typical targets range from about −23 to −14 LUFS. Mastering that hot risks clipping, listener fatigue and rejection during QC against a delivery spec.

Is 100 decibels twice as loud as 90?

Perceived loudness does not scale linearly with decibels, so a decibel increase generally reads as roughly twice as loud to human hearing rather than corresponding to a literal doubling of sound pressure. LUFS measurement accounts for this perceptual curve, which is part of why it differs from a simple peak reading.

What is a good loudness range for LUFS?

There is no single universal loudness range figure that applies to every voice-over, since the right integrated LUFS target depends entirely on the delivery destination, whether broadcast, streaming or another platform. Following EBU R128 or the relevant ATSC guidance for your specific delivery path is more reliable than aiming at a generic number.

Recommended

• Voice Over Pricing
• Hire Voice Actors — Professional Voice Over Artists
• AI Voice Over vs Human Voice Actor — Compare and Choose