What Are the Best Podcast Loudness Targets for Consistent Audio?

Podcast Loudness Targets: The Direct Answer

For most professionally produced podcasts distributed as separate MP3, M4A, or WAV episodes, the best general-purpose target is -16 LUFS integrated loudness for stereo speech, with -19 LUFS for mono speech. These are practical podcast delivery targets, not universal laws of audio. They usually make speech clear and present without forcing every episode to sound louder than a competitive, heavily mastered music release. The target should be applied to the complete episode, including music, sound effects, advertisements, and spoken segments, rather than to isolated dialogue clips.

Also worth reading: How Do AI Mastering Tools Handle Loudness Targets Without Making Everything Sound Too Loud? · What Is Podcast Audio Mastering and How Should Creators Do It in 2026? · What Does AI Podcast Audio Enhancement Actually Do in 2026?

A good companion limit is a true peak of no more than -1 dBTP before lossy encoding. True peak is a more reliable indicator of inter-sample overs risk than the older “sample peak” measurement, and a conservative ceiling helps prevent clipping after MP3 or AAC encoding. Loudness should not be confused with peak level: a waveform can be quiet at its highest sample while still sounding loud, or a waveform can remain below 0 dBFS while containing high inter-sample peaks that distort on some receivers. The widely used -23 LUFS target associated with EBU R128 is appropriate for broadcast-style delivery, but it is not usually necessary for an online podcast mastered specifically for common podcast platforms.

LUFS, Peaks, and Dynamics: What the Numbers Mean

LUFS stands for Loudness Units Full Scale and estimates perceived loudness over time, with greater weight given to frequencies that human hearing notices more readily. A podcast’s integrated loudness is the gated loudness measurement for the full program, excluding some silent or irrelevant material under the measurement standard. This is different from momentary loudness, which changes over short windows, and short-term loudness, which follows broader passages. For episode-level podcast mastering, integrated loudness is usually the primary number because listeners compare one finished episode with the next.

Loudness range, commonly reported as LRA, describes how much levels vary over time. It is not a quality score. A tightly controlled drama, lecture, or narrative podcast may have an LRA around 3–6 LU, while a documentary with natural variation may be wider. The right amount of dynamics depends on the format. Excessive compression can make quiet phrases jump forward, remove intentional pauses, and produce a fatiguing wall of sound. Conversely, an extreme dynamic range may cause quiet sections to disappear on a phone speaker or in a noisy commute. The target should preserve natural speech while ensuring that ordinary dialogue remains consistently intelligible.

The recommendation to use -1 dBTP does not mean every file must reach exactly -1 dBTP. It is a safety ceiling, not a loudness instruction. A properly normalized program may have a lower true peak if it contains a sudden transient, while a quiet program can sit below the ceiling. Measure both loudness and peaks after all processing, check several representative passages by ear, and compare the result with episodes already published by the same show.

Platform Targets and Why They Differ

Podcast platforms and streaming services do not all use the same delivery reference. Apple Podcasts commonly recommends approximately -16 LUFS for stereo and -19 LUFS for mono, with a maximum true peak around -1 dBTP. Spotify’s audio normalization reference has traditionally been around -14 LUFS, and YouTube’s target has also commonly been cited near -14 LUFS. These figures are useful for understanding how music and video services behave, but they do not automatically dictate the right master for a spoken-word podcast. A service may normalize a delivered file, yet it cannot recover clarity, clean dialogue, or detail that was damaged before upload.

The difference between -14 and -16 LUFS is only 2 LU, but perceived loudness changes are logarithmic, so a 2 LU difference is plainly measurable without sounding wildly different. The larger question is whether a show should optimize for a platform’s internal normalization or for the listener’s first encounter with the episode. For a spoken podcast exported through hosting and RSS workflows, a -16 LUFS stereo master is usually easier to control and often lands close to the desired experience without unnecessary over-compression.

How to Set Loudness Without Ruining the Recording

The first step is to establish a clean source. Record at a healthy level with enough headroom, avoid clipping, keep the microphone consistent, and speak naturally. Gain staging matters more than chasing a precise number afterward: a voice recorded several decibels below the available headroom can be raised cleanly, while a clipped recording contains missing information that no normalizer can reconstruct. For a typical voice, peaks around -12 to -6 dBFS often provide a practical recording margin, although the exact value depends on the microphone, preamp, room, and voice.

After editing, remove obvious mouth clicks, plosives, hum, hiss, room noise, and disruptive silence. Apply corrective EQ and de-essing only where needed, then use compression to control inconsistent word-to-word levels. A moderate ratio around 2:1 to 3:1 is often more appropriate for natural speech than aggressive settings, but there is no mandatory ratio. Listen in mono, on headphones, and through a phone speaker because many listeners hear podcasts in those conditions. Finally, normalize the finished program to the chosen integrated target and confirm that the true peak remains below the selected ceiling.

Measure the final file, not just the project timeline. If the show includes a dynamic intro, the LUFS measurement can shift depending on its length and gating behavior. Ads or licensed music may also affect the result. For consistent episodes, decide whether the target includes paid advertisements and whether different feeds or versions of the same show are expected to have identical loudness. A sensible workflow is to keep the same chain, meter settings, and export format across episodes, then document any intentional exceptions.

Compression, Normalization, and AI Enhancement Compared

Normalization primarily changes overall level based on measured loudness. It does not identify a voice, remove echo, repair a noisy room, or decide which frequencies sound muddy. Compression and expansion are more involved: compression reduces louder portions, while expansion or downward expansion reduces the prominence of quieter passages. Together they can improve consistency, but they can also expose pumping, breathing, or unnatural transitions. A mastering chain that reaches -16 LUFS through extreme compression is technically compliant and creatively poor.

AI audio tools can assist with cleanup, speech enhancement, noise reduction, de-reverb, voice isolation, and—in some products—speech or music generation. These features can save time, especially for dialogue recorded in imperfect rooms, but “AI enhancement” is not a single standardized process. One model may preserve a natural voice while another may over-smooth consonants or alter the apparent pitch. Test tools on speech from different speakers, accents, ages, and recording conditions, and compare the processed result with the original.

For Audobox-style AI audio workflows, the sensible use of AI is as an additional editing stage, not as a substitute for metering or listening. Enhancement should precede final loudness normalization so the tool is not asked to repair a level problem. If a creator’s main need is simply “make this episode louder,” ordinary normalization and light dynamic control are cheaper and more predictable than an AI redesign of the voice.

Common Loudness Mistakes and How to Avoid Them

One common mistake is targeting the loudest moment instead of the integrated program. A single shout, music sting, or advertisement can cause a limiter to pull down the rest of the episode. Another is using only a peak meter. A file may peak at -3 dBFS and still be much quieter in perceived terms than another file with the same peak, or it may contain inter-sample peaks that are not obvious on a basic meter. Another mistake is exporting different settings for the RSS file, YouTube version, trailer, and social clips without labeling them.

Over-compression is the most audible format-wide error. If every syllable has identical level, breaths disappear, quiet emphasis is flattened, and the narrator may sound unnatural. Raising the entire track after heavy compression can also make the noise floor more obvious. Do not compare a mastered podcast with a raw voice recording and assume the difference is entirely caused by loudness. A better comparison is between two final masters from the same show, both exported from the same processing chain.

Finally, do not assume a stricter number always produces a better result. Targeting -14 LUFS is not automatically “more professional” than -16, and targeting -23 is not automatically bad. The correct target depends on the distribution system, content, artistic direction, and listener environment. Measure after export, inspect the report, listen at normal volume, and make small changes. A two-LU deviation is usually less important than distorted speech, an abrupt edit, or a soundtrack that overwhelms the host.

When to Act, and What It May Cost

Act on loudness when publishing regularly, receiving complaints about episodes being too quiet or too loud, changing microphones or recording rooms, adding music, or moving from raw recordings to a finished master. There is little reason to reprocess every archive episode merely to satisfy a new preference if the existing files are consistent and listen well. A new show should establish a reference episode, measure it, and use that reference when evaluating later work.

The cost of proper loudness mastering is often lower than the cost of rebuilding bad recordings. A free or bundled audio editor may provide LUFS and true-peak meters, while a dedicated podcast editor may offer episode templates, batch export, dynamic processing, and integrated hosting support. AI cleanup products commonly use subscription pricing, usage credits, or plan-based limits, and prices change frequently enough that a fixed universal price would be misleading. The relevant cost comparison is not simply “free versus paid”; it is whether the tool saves editing time without changing the speaker’s identity or introducing artifacts.

For a small independent show, a practical starter setup is a reliable recorder, a straightforward editor with loudness metering, a documented -16 LUFS stereo or -19 LUFS mono target, and a -1 dBTP ceiling. Spend first on room treatment, microphone placement, and editing discipline. Consider AI enhancement when noise or room problems are the bottleneck, and consider professional mastering when the show has advertising, major sponsorship, or a strong requirement for consistent delivery across venues and platforms. The best target is the one that delivers intelligible, natural speech, passes encoding tests, and remains consistent from one episode to the next.