Loudness Targets for Dubbed Video: LUFS Explained Simply
When you dub a video, you'll notice something almost immediately: the dubbed audio sits quieter than the original, even when you haven't touched anything else. This happens because TTS synthesis and voice cloning produce audio at a different loudness level than natural speech. Before you publish, you need to normalize it to match platform expectations and your original mix.
That's where LUFS comes in. LUFS stands for Loudness Units relative to Full Scale. It's the standard measurement platforms use to keep audio consistent across all videos and content. Understanding it takes the guesswork out of matching your dubbed track to the rest of your mix.
What LUFS actually measures
LUFS isn't the same as the raw decibel level you see on a meter. It accounts for how human ears perceive loudness. A signal at -23 LUFS will sound roughly as loud as another at -23 LUFS, regardless of whether that signal is speech, music, or noise. Platforms use LUFS because it gives them a consistent, perceptually fair target across millions of videos.
When you export dubbed audio, most tools default to -18 to -20 LUFS. Natural dialogue sits around -18 to -16 LUFS. The TTS voice often comes out quieter, around -22 to -25 LUFS, because the algorithms that generate speech are conservative about peak levels.
The key difference between LUFS and simple peak metering matters for dubbing. A piece of content might have quiet dialogue followed by sudden music swells. Peak meters will spike at the music; LUFS integrates the loudness over time and weights it according to human perception. This is why two audio files can have the same peak level but very different LUFS measurements. LUFS is also what matters for automatic normalization on platforms, so it's the only number you should care about before publishing.
Platform loudness targets
Different platforms have different requirements, and choosing the wrong target will either force the platform to adjust your audio automatically (introducing artifacts) or leave it sounding out of place compared to other content.
YouTube uses -14 LUFS as its standard. If your video audio sits at -18 LUFS, YouTube will automatically normalize it up to -14, which can introduce distortion or sound artifacts depending on your headroom. This happens during upload without your control.
Podcast platforms (Spotify, Apple Podcasts, most podcast hosts) target around -16 LUFS for speech. Podcasts are listened to on mobile devices, earbuds, and in noisy environments, so a slightly hotter level helps the voice cut through. If your podcast audio is too quiet, listeners will turn up their volume and then get startled when music interlude plays.
Streaming video (Netflix, Prime Video, Vimeo) typically specifies -27 to -24 LUFS for dialogue. These platforms compress the dynamic range significantly to protect viewers from unexpected loud passages, so they accept quieter input because they'll process it anyway.
Social media (TikTok, Instagram, Facebook) is less strict and usually accepts anything between -14 and -20 LUFS without complaint, but consistency still matters for user experience. Videos that sound noticeably quieter than surrounding content lose engagement.
Always check the platform you're uploading to. The documentation is usually one search away, often under audio specifications or video upload guidelines.
Why dubbed audio comes out quieter
Synthetic speech avoids peaks. Human speech has natural breath, emphasis, and dynamic range. When a voice actor says a line, they naturally stress certain words, which pushes the peak level higher. TTS approximates this, but the algorithms err on the side of caution to avoid distortion in the generated waveform.
Additionally, if you're using voice cloning from a reference recording, the output inherits some of the quietness characteristics of the original voice sample. A softer reference voice will produce softer synthesized output. If your original speaker has a naturally quiet delivery, the dubbed version will tend toward quiet output as well.
Some dubbing workflows separate the dubbed voice and the background music/ambiance into different tracks during processing. If you're not recombining them before measuring, you're looking at the voice track in isolation, which will naturally measure quieter than the full mix. The background elements add perceived loudness, so when you measure the voice alone, you see only part of the picture.
Finally, compression settings during synthesis matter. Some TTS engines apply lookahead compression to prevent clipping, which reduces overall level. The safer the compression, the quieter the output.
Choosing your target: a decision framework
Before you normalize anything, decide which platform matters most for this video. If it's going to multiple platforms, pick the one with the most restrictive requirement as your primary target, then adjust afterwards if needed.
Ask yourself: Is this video primarily for YouTube? If yes, aim for -14 LUFS. You won't need post-normalization adjustment, and YouTube's algorithm treats properly normalized content fairly.
Is it a podcast or audio-first content? Aim for -16 LUFS. This level works across all major podcast platforms and sounds natural on phone speakers and earbuds.
Is it going to Netflix, Prime, or other streaming services? Check their specific requirements first. Most want -27 to -24 LUFS for dialogue. Normalizing higher will force their player to turn it down, which sometimes triggers artifacts.
Is it social media only? You can go as high as -14 LUFS without problems, but -16 to -18 LUFS is safer because it sounds more natural and less compressed across different playback devices.
Once you've chosen, write down your target number. You'll need it for the next step.
Measuring your own audio
Before you normalize, you need to measure your current LUFS. Free tools can do this reliably.
Audacity, the free waveform editor, has a loudness analysis feature. Open your exported dubbed file, select all audio, go to Analyze menu, find Loudness, and it will display your integrated loudness in LUFS. The measurement takes three to five seconds.
Professional DAWs (DaVinci Resolve, Premiere Pro, Logic Pro) have built-in loudness meters that display LUFS in real time. If you're already editing in one of these, use the meter before exporting. Most DAWs also let you preview what normalization will do without committing to it.
Online LUFS meters exist, but don't upload sensitive or confidential content to them. Stick to free desktop tools for anything you care about.
The measurement only takes once per project. Write down your current LUFS and your target LUFS. The difference between them tells you how much adjustment you need.
Normalizing your audio
Once you know your current LUFS and target, you have two paths: use ffmpeg from the command line, or use a DAW if you're already editing there.
Using ffmpeg
Here's the basic command for command-line normalization:
ffmpeg -i input.mp4 -af loudnorm=I=-14:TP=-1.5:LRA=11 output.mp4
The parameters mean:
I=-14is your target integrated loudness in LUFS. Replace -14 with your target (-16 for podcasts, -24 for Netflix, etc.).TP=-1.5is the true peak limit, preventing distortion from clipping. This is a safety ceiling; audio won't exceed this peak level.LRA=11is the loudness range, a measure of dynamic variation. Twelve is the standard ATSC value; adjust downward only if you want the output to sound more compressed.
For a podcast or speech-heavy content, use -16 instead of -14. For streaming platforms, use their specified target.
Run it once and measure the output with your LUFS meter to verify. If your DAW shows the normalized file is now at your target, you're done. If it's still off, you can iterate by adjusting the I parameter.
Using a DAW
If you're already in Premiere Pro, DaVinci Resolve, or Logic Pro, use the built-in loudness meters and correction. These tools usually let you set a target LUFS and preview the result in real time, then export directly.
Open your final mixed file, enable the loudness meter, and watch the integrated loudness number as you play through. If it's below your target, most DAWs have a loudness correction tool that will adjust the audio to meet that target. Apply it, and then export your final file.
Common mistakes and how to fix them
Normalizing before your final mix. If you normalize the dubbed voice track before adding music and effects, you'll have to normalize again after the mix, which can introduce cumulative distortion. Always normalize after the final mix is complete.
Measuring only the voice track, not the full mix. If your dubbing workflow produces separate voice and background tracks, measure them combined as they'll appear in the final video. Measuring the voice alone gives you an artificially low LUFS reading.
Using the wrong target for your platform. If you aim for -14 LUFS but upload to a Netflix-style platform that wants -24, your video will sound loud and compressed compared to other content on that platform. Always check platform requirements first.
Not accounting for headroom. If your audio already has significant headroom before normalization, the loudnorm filter will push it up. If you're starting close to clipping, pushing harder will introduce distortion. Always measure first and only normalize if you have at least 3 dB of headroom.
Forgetting that normalization is loudness, not volume. Turning up volume in post-production is a quick fix, but it's not the same as correct loudness normalization. Loudness normalization accounts for perceptual loudness and prevents the platform from re-normalizing. Volume changes are raw amplitude adjustments. Always use loudness normalization, not just volume boost.
Worked example: dubbing a YouTube travel vlog
Let's walk through a concrete scenario. You've recorded a 12-minute travel vlog in English, with natural dialogue, ambient sounds, and music. You dub the dialogue into Spanish, and now you need to prepare it for YouTube.
First, export your final mix with the dubbed Spanish voice, background ambiance, and music combined. Measure it with Audacity. It shows -18.5 LUFS. YouTube's standard is -14 LUFS, so you need to bring it up 4.5 dB.
Run the ffmpeg command with I=-14. The command processes the file and outputs a new version. Measure the output again. It now reads -13.8 LUFS, which is close enough. Export that version and upload to YouTube. The platform recognizes the correct loudness level and doesn't apply automatic normalization, so your audio plays at the intended level without artifacts.
If you'd uploaded the -18.5 LUFS version without normalizing, YouTube would normalize it up to -14 LUFS automatically. For this particular audio, that might have been fine, but if your original mix had tight headroom near the peaks, the automatic normalization might have introduced slight clipping or artifacts that you never intended.
Understanding LUFS targets by content type
Different content types benefit from different LUFS strategies. Understanding why each category has its target helps you make the right choice for your specific video.
Dialogue and educational content
Educational videos, interviews, and narrative-driven content rely heavily on intelligible speech. These work best at -14 to -16 LUFS because the viewer expects clarity. If the dialogue is too quiet, the viewer will fiddle with their volume, which becomes a usability problem. Dialogue-first content should always be normalized tight to its platform standard.
Music-forward content
Videos where music is the primary focus (music videos, performances, DJ sets) can sit at -18 to -20 LUFS safely. The music itself will have dynamic range that prevents constant loudness, and viewers expect musical content to have more dynamic character than dialogue. Pushing music-heavy content to -14 LUFS can make it sound over-compressed and fatiguing.
Mixed content (dialogue plus music)
Travel vlogs, product reviews, and lifestyle videos have both speech and music. Normalize based on whether dialogue or music is the focus. If it's mostly talking with background music, aim for your platform's dialogue standard. If music is equally important, aim slightly lower to preserve the dynamic range of both elements.
Ambient and atmospheric content
ASMR, nature sounds, meditation, and similar content should stay quieter, usually -20 to -24 LUFS. These categories depend on the listener actively choosing to play quiet content in a controlled environment. Over-normalizing defeats the purpose.
Troubleshooting: when normalization doesn't work
If you normalize your audio but it still sounds wrong, the issue usually isn't LUFS. Common problems include:
The normalized file sounds distorted or grainy. This happens when you start with audio that's already near clipping. Normalization can't add headroom; it can only adjust what's there. If your dubbed track has peaks already close to 0 dB, pushing it harder will clip. Solution: if you generated the dubbed audio yourself, try regenerating the voice at a lower input level if your TTS tool allows it. If you're working with an exported file, you may need to apply gentle compression or limiting before normalizing to create headroom.
The normalized file is still quieter than expected. Confirm you're measuring the complete mix. If you measured only the voice track while your final mix includes music and ambiance, the reading will be artificially low. Measure the combined export, not individual tracks.
Peaks are still too loud after normalization. The true peak limit (TP parameter) should prevent clipping, but some ffmpeg configurations don't apply it correctly. Try setting TP to -3 instead of -1.5, which provides more margin. Alternatively, use a DAW with proper metering to ensure the true peak stays within bounds.
The audio sounds too compressed or lacks dynamics. LUFS normalization shouldn't remove dynamics; it just shifts the average loudness. If the audio sounds compressed, something else in your mix needs adjustment. Common causes include: the TTS voice itself being naturally compressed, the original voice sample being over-compressed, or a previous normalization pass having already squeezed the dynamics. If this is the case, check if your TTS has quality or compression settings you can adjust before regenerating.
When to normalize and when to skip it
Normalization is usually necessary, but not always. Ask yourself: Does my audio need adjustment?
If your dubbed audio already measures within 1 dB of your platform target, skip normalization. Don't normalize for the sake of it. Every processing step carries the risk of introducing subtle artifacts, so only do it when the audio genuinely needs it.
If your audio is more than 1.5 dB away from your target, normalize. The difference will be audible to viewers, especially on mobile devices or lower-end speakers.
If you're uploading to multiple platforms with different targets, normalize to the most restrictive standard first (usually -14 to -16 LUFS). Platforms that accept quieter audio will still accept the normalized version; platforms that reject loud audio will prefer the normalized version.
When and how to check your work
After you normalize and export, verify the result before publishing. Open the file one more time in your measurement tool and confirm the LUFS is within 0.5 dB of your target. If it's off by more than 1 dB, re-normalize with adjusted parameters.
Listen to the normalized file on the actual playback device you expect viewers to use. If it's a YouTube video, watch it on YouTube after publishing (unlisted, if you want privacy) and listen on your phone, laptop, and headphones. The goal is that the dubbed audio sounds as natural as the original version would have, just in a different language.
If the normalized audio sounds compressed, robotic, or fatiguing to listen to, the problem isn't LUFS; something else in your mix needs adjustment. LUFS normalization shouldn't change the character of the audio, only its level.
The verification step is easy to skip but worth the time. A few minutes of listening catches problems that measurements alone won't reveal. Your viewers are listening on wildly different devices and headphones; what sounds right on your studio monitors might need adjustment for the majority of viewers in earbuds or phone speakers.
What to do next
Before you dub your next video, identify your primary distribution platform and find its audio specifications. Write down the target LUFS. After you finish your final mix, spend five minutes measuring and normalizing. Export the corrected version, spot-check it on a real device, and publish with confidence.
DubLab handles the voice synthesis part and produces audio at a consistent output level, but the loudness stage is yours to control. Spending a few minutes on this step separates videos that sound polished and professional from ones that sound like an afterthought. The time investment is minimal; the difference in perception is everything.
🚀 Start Dubbing Your Videos Today
DubLab uses AI to translate your videos into 92+ languages in minutes.