For a speech-led 24/7 kids’ story stream, measure each complete episode and use −18 LUFS integrated as a practical starting point for consistent distribution. Then check true peak on the processed, encoded file and listen across episode boundaries; the number is not a child-specific safety limit.
Normalisation can reduce jarring differences between a quiet narration and a louder next episode, but it cannot fix every mix problem. The result depends on the full programme, its music and effects, the processing you apply, and the way the stream is encoded and played.
Why episode loudness varies
Episodes often come from different recording sessions, microphones, rooms or editors. One narrator may speak close to the microphone; another may sit further back. A story recorded in a quiet room can have clear, even speech, while an older recording may contain room noise or a softer performance. Peak readings alone do not tell you whether those episodes will sound equally loud to a listener.
Integrated loudness estimates perceived programme loudness over a defined piece of audio. LUFS is an absolute loudness scale relative to full scale, and ITU-R BS.1770 specifies methods for measuring programme loudness and true peak. A loudness meter therefore answers a different question from a meter that reports only the highest sample value: it helps compare the overall level of one complete programme with another.
The contents matter too. A quiet story with pauses, soft dialogue and little music may have a low integrated reading but still contain an unusually loud sound effect. A programme with continuous music can measure differently from one led almost entirely by speech. Background beds, opening titles, closing credits and silence all contribute to the measured episode, so decide what counts as the programme and measure it consistently.
Treat the listener’s experience as the goal, not a matching set of numbers for its own sake. You want dialogue to remain intelligible and episode changes to feel controlled while preserving the intended pauses, emphasis and contrast. If your channel also uses a continuous devotional playlist, the practical issues of building a repeating programme are covered in how to run a continuous bhajan playlist as a YouTube live stream, though its music-led content needs its own listening checks.
Measure each complete episode
Start with the rendered episode file that you intend to put into the catalogue. Use one loudness meter that measures integrated LUFS or LKFS using a BS.1770-compliant method, and record both integrated loudness and true peak. A meter’s documentation should make clear which method it uses; without a consistent method, readings from different tools may not be directly comparable.
Measure the whole episode, from the opening sound you will actually stream through the end of its closing material. Do not measure just a paragraph of narration or a convenient loud section. A short sample may miss a quiet introduction, a music cue, a long pause or a sharp effect. Keep the programme boundaries stable across episodes. If the stream adds a separate channel ident between files, document that separately rather than including it in only some episode measurements.
A simple record can include the asset name, source integrated loudness, source true peak, final integrated loudness, final true peak, gain applied, any limiting, and the meter or measurement method. The record is useful when an episode sounds different later: you can distinguish an unusually quiet source from a processing change or a different export. For a large catalogue, a spreadsheet is enough to begin; the important part is recording the same fields every time.
Use the same meter settings and measurement mode for all episodes. If you are choosing a tool, verify that it reports integrated loudness and true peak, rather than only peak sample level or short-term loudness. Make a test with a few contrasting episodes: one mostly spoken, one with music or effects, and one you already know sounds quieter or louder. The numbers help identify a starting adjustment, but they do not replace listening.
The audio file is one part of the eventual stream chain. If you are preparing long programme files, the practical considerations in how to encode a long video overnight without filling a PC drive can help you plan the export stage; do not assume the export itself makes the audio levels consistent.
Use −18 LUFS integrated as a starting point
For measurable speech, the Audio Engineering Society’s internet-audio distribution guidance gives −18 LUFS as a recommendation and includes drama at that level in radio-style stream examples. That makes −18 LUFS integrated per complete episode a defensible starting point for a narration-led story stream. It is distribution guidance for this kind of speech-led material, not a special rule for children’s content.
No child-specific loudness target is established by the guidance discussed here. Do not interpret −18 LUFS as a hearing-safety level, a guarantee of comfortable listening, or a substitute for sensible playback-volume choices. Parents and carers still control the listening environment and device volume; the production target is about making episodes more consistent when distributed.
Other published targets belong to different contexts. EBU R 128 recommends −23 LUFS for broadcast programmes. Apple’s guidance for podcast audio preconditioning discusses around −16 dB LKFS with a ±1 dB tolerance, alongside a true-peak recommendation. Neither should be lifted automatically as the target for a continuous YouTube stream. If you are making a separate broadcast or Apple Podcasts deliverable, follow the relevant current official guidance for that destination.
| Guidance | Published figure | Intended context | How to use it here |
|---|---|---|---|
| AES AESTD1008 | −18 LUFS for measurable speech | Internet audio distribution; includes drama examples | A reasonable starting point for narration-led episodes, then validate in your own stream |
| EBU R 128 | −23 LUFS | Broadcast programme loudness | Useful context, not an automatic target for an internet stream |
| Apple Podcasts for Creators | Around −16 dB LKFS ±1 dB | Podcast audio preconditioning and Apple playback | Use for that delivery context, not as a universal continuous-stream target |
The figures are not interchangeable just because they all describe loudness. They reflect different delivery contexts and assumptions. AES also notes that speech and music normalised to the same integrated loudness can be perceived differently; a mixed programme may therefore need careful balance even when its overall reading matches the target. A story stream with a prominent music bed should be judged as a complete listening experience, not tuned by number alone.
Read the current source material before adopting a target for another destination: the AES internet-audio recommendations, EBU R 128, and Apple Podcasts audio requirements describe their respective contexts. Standards and platform documents can change; check the relevant official page when you prepare a separate deliverable.
Apply consistent gain across episodes
Once you have a baseline, compare each episode’s integrated reading with the chosen target. Gain adjustment is the simplest first move: if an episode is below the target, raise its level; if it is above, lower it. Loudness normalisation is fundamentally an adjustment to match programme loudness, not an instruction to compress every episode until its quiet and loud moments are identical.
A quieter source can need substantial gain to reach the target. Before applying it, check what happens to the true peak and to background noise. Increasing the whole episode raises narration, music, room tone and effects together. If a hiss or room sound becomes distracting, gain alone will not solve it; revisit the recording or mix if possible rather than treating the meter as a repair tool.
If reaching the target would create peaks that are too high for the next stage, a limiter may control them, but use it only where needed. Limiting can alter transients, make effects sound harder, or change the character of a performance. Listen to the result and compare it with the source. If the processing is audible, consider a lower final loudness for that episode or a better mix rather than forcing the same figure at any cost.
Keep the adjustment predictable. A written rule might say: measure the complete episode, set gain toward −18 LUFS integrated, use limiting only if the gain would create a peak problem, then re-measure the final encoded file. Do not make ad hoc boosts based on a single quiet sentence. If the narration is hard to hear only because music masks it, adjust the relative music-to-voice balance; raising the whole file will not change that relationship.
This distinction is especially useful for creators who schedule recorded material rather than speak live. The workflow for scheduling recorded news videos in English and Hindi on YouTube Live in India is a different editorial use, but it shares the need to prepare and check each asset before it becomes part of a continuous channel.
Check true peak after processing and encoding
Integrated loudness and true peak are complementary checks. Integrated loudness describes the programme’s overall measured level; true peak estimates the highest reconstructed peak between digital samples. A file can have an appropriate integrated reading and still have a peak that clips or causes trouble later in the chain. Conversely, a low peak does not prove the programme is loud enough or consistent with its neighbours.
After gain adjustment or limiting, measure both values again. Then encode the actual delivery file and measure that file too, rather than relying only on the production master. Lossy encoding can change peak behaviour, and downstream encoders or players may handle audio differently. Your final check should be on the version that will be used in the stream, not just the project export before compression.
Apple’s −1 dB FS true-peak recommendation is specific to its podcast preconditioning guidance; it is not a universal rule for every YouTube stream. Use it only when preparing that Apple Podcasts context, and check the requirements for the destination you actually serve. For your continuous stream, leave appropriate headroom for the encoder and playback chain, and use the platform and encoder guidance that applies to your configuration.
If the file has loudness metadata, it may help a compatible player adjust playback gain, but not every streaming path honours that metadata. Do not rely on metadata as a substitute for preparing consistent audio. Recheck a sample through the same encoding and streaming path you plan to use, and compare what the meter says with what you hear.
Listen to episode transitions
A measurement pass can make files comparable; the transition listen tells you whether the result works. Put adjacent episodes in sequence, including the endings and openings as they will actually appear in the stream. Listen at a representative playback level and check whether the first narration after a pause is clear, whether an opening sting is disproportionately loud, and whether music suddenly becomes dominant or disappears.
Pay special attention to a quiet episode followed by a lively one, and the reverse. Sudden level changes can be noticeable even when both files measure close to the target, because their arrangements differ. A story with soft narration and a bright music bed may feel louder than a drier episode with the same integrated reading. Compare speech intelligibility, music balance, effects, pauses and the change in perceived level across the boundary.
Listen to a short section from each episode as well as the actual join. A full-episode average can conceal an opening that is too quiet or a loud effect in the middle. If one section feels wrong, note the time and the cause before changing the whole file. A local mix adjustment may be more appropriate than a global gain change, which affects every moment equally.
Use more than one ordinary listening situation if practical, such as a television or speaker at a modest setting and a phone speaker. This is not a laboratory test and does not establish a safety threshold; it is a check that dialogue and transitions remain understandable on the devices your audience is likely to use. Avoid turning the test into a search for one perfect playback volume, since listeners have different devices and environments.
Keep the method consistent across the catalogue
Write down the target, measurement method, episode boundaries, export format and processing rule. When you add a new story, follow the same steps and save its readings. If you revise the target later, decide whether older episodes need reprocessing; changing only new uploads can create a new mismatch between catalogue generations.
For older files, prioritise episodes that are known to be unusually quiet, harsh or inconsistent, then check them against their likely neighbours. Avoid reprocessing everything merely because a meter exists. Keep original masters where possible so you can revisit a mix without compounding previous processing, and label final files clearly enough that the stream uses the checked version.
If the programme includes songs, extended ambience or sound effects, do not assume the speech-led target settles every balance question. The AES guidance notes that equal integrated loudness does not always produce equal perceived loudness between speech and music. Preserve intentional dynamics and make a separate editorial decision about the relative level of beds and effects. Consistency means avoiding accidental jumps, not making every moment equally loud.
Your workflow also has an operational side: the prepared episode must reach the stream in sequence, and a long-running channel needs a reliable hand-off between files. StreamNeo is relevant when the particular pain is keeping a prepared video playing after your own computer is switched off; it lets you upload the file and run the YouTube broadcast without installing software locally. Audio preparation still belongs in your own workflow, and you should verify the final stream as well as the files.
Before committing, compare the operating options on the pricing page. When the file and channel are ready, start free — 24-hour trial, no card.
FAQ
Is −18 LUFS a safe volume for children?
No. It is a distribution starting point for measurable speech, not a child-specific safety limit or a guarantee about listening exposure. The actual sound level at the listener depends on the playback device, volume setting and environment.
Should I use −23 LUFS or Apple’s podcast target instead?
Not automatically. EBU R 128’s −23 LUFS figure is broadcast guidance, while Apple’s approximate −16 dB LKFS recommendation applies to podcast audio preconditioning. Use the guidance for the destination you are preparing, and do not treat these figures as equivalent targets for a continuous YouTube stream.
Do I need to make every episode exactly −18 LUFS?
Use it as a consistent starting point, not a reason to damage a mix. If reaching it creates excessive peaks, audible limiting or distracting background noise, make a considered adjustment and listen to the result rather than forcing the number.
Is measuring the master file enough?
It is a useful first check, but measure the final encoded file as well because encoding can affect peak behaviour. Then listen across episode joins in the actual stream path, since neither a loudness reading nor a true-peak value alone tells you whether the transitions sound natural.