For a 24/7 YouTube podcast stream, make episodes sound similar in perceived loudness, rather than simply making their peak readings match. Measure programme loudness, adjust each episode before playout where practical, check true peaks, and listen to the joins between episodes.
YouTube does not publish a specific LUFS target for Live in its encoder guidance. EBU’s −23 LUFS recommendation is a broadcast reference, not a YouTube requirement; choose a consistent approach for your material and confirm it by listening to the actual stream.
Measure the whole programme, not just peaks
A peak meter answers a narrow question: how high did the signal rise at a moment? It does not tell you whether a complete episode sounds louder or quieter than another. A close-miked, gently spoken conversation and a lively interview can reach similar peaks while having quite different average loudness. Matching those peaks alone can therefore leave your playlist uneven.
Use a loudness meter that reports integrated or programme loudness in LUFS. Measure complete episodes where feasible, or a representative section long enough to include the kinds of speech, music and pauses that occur throughout the programme. A short isolated sentence is a poor basis for adjusting an hour-long conversation, because delivery varies over time.
Make a small record of readings for episodes, recurring introductions, adverts and music beds. The purpose is comparison: if one episode measures substantially differently from the rest, it is a candidate for listening and adjustment. A measurement is not a verdict on whether the content sounds right. Two episodes with the same integrated reading may still differ in timbre, speech density, or the prominence of music.
Do not use an ordinary peak display as a substitute for LUFS measurement. OBS documents its meter as a sample-peak programme meter; its VU-style display is an RMS reading over 300 ms, not an integrated loudness meter. That is useful for watching signal level and avoiding clipping, but it cannot establish programme loudness. See OBS’s notes on audio clipping and its meters.
The European Broadcasting Union’s EBU R 128 version 5 recommends average programme loudness of −23 LUFS and the use of Loudness Range and Maximum True Peak Level descriptors. That gives you a standards-based reference if you need one, not an automatic preset for a YouTube channel. The EBU’s separate streaming supplement explains that playback conditions matter: a phone in a noisy place is not the same listening context as a capable system in a quiet room. A reference level cannot guarantee the same listener experience everywhere. You can read EBU R 128 and its streaming supplement before deciding whether either reference suits your distribution choices.
Choose offline preparation or live correction
If your episodes are recorded files and can be prepared before broadcast, offline normalisation is usually easier to inspect. You can measure each file, make a gain change, render a corrected copy, and check its peaks before it enters the playlist. If the result is wrong, you can revise it without changing a live broadcast.
Real-time processing is useful when material arrives continuously or you cannot prepare every item beforehand. A controller can respond to changing loudness as the stream plays. But it cannot know what audio comes next, so it cannot plan a smooth correction across an episode boundary. It may also react to intentional changes in a programme, such as a quiet reflection after a lively exchange, and reduce the contrast the producer meant to preserve.
| Approach | Where it helps | Main trade-off | Check before relying on it |
|---|---|---|---|
| Offline normalisation | A known playlist of files that you can prepare ahead | Requires a repeatable preparation step when files change | Episode loudness, output true peak and joins |
| Real-time correction | A changing source or playlist that is impractical to prepare in advance | Can react late, pump, or alter expressive dynamics | Response over time, speech dynamics and sudden level changes |
| Gain only | A file that is consistently too quiet or loud, with enough headroom | Raises or lowers peaks along with the rest of the programme | Whether the corrected signal clips or becomes too quiet |
| Compression and limiting | Material with troublesome level variation or peaks | Excessive processing can sound tiring or flattened | Speech emphasis, music, true peaks and artefacts |
The Audio Engineering Society describes rolling integration periods from 30 seconds to two minutes as contextual guidance for active real-time loudness controllers; longer periods generally sound more natural and can produce fewer audible artefacts. This is not a required setting for YouTube, nor a magic number that suits every show. If you use a live controller, start conservatively and listen for how its response affects actual speech and transitions. The AES loudness normalisation resource discusses the limits of real-time correction.
There is a practical workload trade-off. Offline work takes attention before a file joins the rotation, but its result can be inspected and reused. Live correction shifts some work into broadcast time and can be less predictable when a playlist changes. If your workflow already rotates recorded episodes, the advice on using a radio automation system with YouTube Live may help you think about where file checks belong in the hand-off to playout.
Apply gain and restrained dynamics
Once you have chosen a sensible reference for your channel, adjust gain to bring each programme closer to it. Treat LUFS as a way to compare episodes consistently, not as a promise that every listener will hear the same volume. The EBU’s −23 LUFS reference can be appropriate when you want to align with its broadcast practice. EBU R 128 S2 also describes a conditional interim distribution range of −20 to −16 LUFS where metadata is not managing device gain and a broadcaster chooses to control dynamic treatment before streaming. That context is not a YouTube preset; do not select a number from the range without considering your programme and playback path.
For speech that is simply quieter than the rest of the playlist, a measured gain adjustment may be enough. Gain raises the programme and its peaks together. If that leaves insufficient headroom, do not keep increasing it just to make a meter match. Listen for distortion, then decide whether a gentle compressor or limiter is needed, or whether a lower overall level better preserves the recording.
Compression can reduce the difference between quieter and louder passages. Used lightly, it may make speech easier to follow when a presenter moves away from the microphone or changes delivery. Used heavily, it can make breaths and room noise conspicuous, flatten emphasis, and tire listeners over a long session. A devotional reading, a quiet interview and an energetic discussion may each need different judgement; one aggressive preset is unlikely to respect all three.
A limiter is primarily a way to constrain peaks, not a substitute for balancing whole episodes. If you use one, listen for clipped consonants, harshness or a squeezed sound, especially on laughter, music and emphatic speech. Make changes with the episode as a whole in mind, then remeasure and audition the result. It is often better to accept a modest difference in level than to remove meaningful dynamics merely to make every file show the same reading.
For a non-technical workflow, keep a clean original and make corrected copies rather than overwriting source recordings. Name the outputs clearly, note the processing applied, and use the corrected versions consistently in the playlist. This gives you a way back if a revised normalisation pass sounds worse after a night of listening.
Check true peaks and encoding headroom
Programme loudness and peak control solve different problems. Integrated LUFS helps compare the level of whole programmes; sample peaks and true peaks help reveal brief high points that may clip or become harsher during encoding. An episode can have a similar LUFS reading to its neighbours and still contain a troublesome peak.
A sample-peak meter measures the digital samples in the file. A true-peak meter estimates the signal between samples as well, where reconstruction and lossy encoding can create a higher peak than the sampled values suggest. If your meter provides a maximum true-peak reading, use it as a separate check after gain and dynamics processing. If it does not, at least inspect the output for sample clipping and listen carefully to loud moments; do not mistake a peak display for a full loudness analysis.
Keep the encoded output below 0 dBFS to avoid digital clipping, as OBS’s guidance notes. The exact margin you choose is a workflow decision, not a YouTube loudness rule. Leave enough room for the encoder and avoid pushing a limiter right against the ceiling simply because the signal appears louder that way. If a file clips before it reaches your streaming software, a downstream fader cannot restore the lost waveform.
Check the processed export, not only the source. Gain changes, compression and limiting can change the output peak, and the live encoder adds another stage. If an episode is re-rendered, measure and listen to that version before it goes back into rotation. Keep a record of any recurring file that needs special treatment so you are not rediscovering the same problem after each playlist update.
For YouTube delivery, follow the current YouTube Live encoder guidance for transport settings and test procedures. The page specifies audio format and delivery settings; it does not specify an official LUFS target for Live. Keep that distinction clear when documenting your setup: delivery settings concern how the signal is sent, while loudness decisions concern how the programme is prepared and heard.
Listen across episode transitions
A stream can contain well-prepared individual files and still feel uneven at the joins. The end of one episode may fade into silence, while the next begins with a loud theme. An advertisement, a station ident or a music bed can make the change more obvious. Measure each component that will actually play, then listen to the sequence in the same order as the playlist.
Pay attention to the first moments after a transition. If the outgoing episode ends quietly and the next begins at full level, a listener may hear a sudden jump even when both episodes have reasonable integrated readings. Conversely, a long pause may be intentional in a reflective programme. Do not trim or compress silence automatically without considering editorial intent and the channel’s format.
Make a test sequence using the quietest and loudest episodes, plus recurring material such as intros, adverts and interstitials. Include the precise fades and gaps used in playout. Listen at a normal listening level, rather than turning the volume up to inspect a quiet passage and forgetting to reset it before the next item. Check both the whole sequence and the transition points; the latter are where mismatched files are easiest to notice.
When a join is distracting, first identify its cause. It may be a level mismatch, a sudden change in music, an abrupt edit, or silence of a different length. A gain adjustment may fix a level mismatch; it will not fix a badly timed edit. Keep changes specific to the problem, and recheck the full sequence after adjusting one file so you do not create a new imbalance elsewhere.
If you are running a recorded podcast as a continuous channel, episode order and naming matter alongside audio. The guide to adding episode names to a 24/7 YouTube podcast stream covers the presentation side; use the same playlist order when auditioning its audio, so you hear the transitions your audience will actually encounter.
Test the continuous stream, not only the files
A file that sounds correct in an editor may behave differently once sent through the encoder and heard from YouTube. Test with audio similar to the material you plan to broadcast: ordinary speech, the loudest passages, music, and the quiet sections that occur in the real playlist. YouTube recommends testing with activity similar to the planned stream and monitoring stream health. Its guidance is about delivery and monitoring rather than a loudness target.
Listen to the actual live output on a device and connection representative of your audience where practical. Check that the signal is present, speech remains intelligible, music does not dominate unexpectedly, and there is no audible clipping or pumping. Compare episode changes without constantly adjusting the playback volume. If you need to raise or lower the volume at every boundary, revisit your preparation rather than treating the listener’s volume control as the normalisation system.
A continuous channel needs a repeatable process because the playlist can change after the first test. When an episode, intro, advert or source filter changes, make a short check of the revised sequence and verify the outgoing stream again. Keep a note of the date and what changed, so that if levels shift overnight you can narrow the cause to a new file, a changed source level or an encoder adjustment.
Do not treat an automated restart as an audio check. If an unattended stream drops, it may restart successfully while a new source or changed playlist still has the wrong level. If you rely on a local computer for playout, keep power and sleep settings in the test plan; this guide on preventing OBS from sleeping during a 24/7 stream addresses that separate continuity risk. Audio consistency and stream continuity need their own checks.
If file preparation is the part most likely to be skipped between episodes, a workflow that lets you upload a video once and keep the broadcast running while your computer is off can remove that specific repeat-playout burden. StreamNeo is useful in that situation, but it does not choose your loudness reference or replace measuring, processing and listening to your audio before upload.
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
What LUFS should I use for a YouTube live stream?
YouTube’s Live encoder guidance does not publish a required LUFS target. EBU’s −23 LUFS recommendation is a broadcast reference, not a YouTube rule; choose a consistent level for your programme and confirm it by listening to the stream.
Is matching peaks enough to make podcast episodes sound equally loud?
No. Peak readings describe brief high points, while integrated loudness helps compare the overall programme. Use both checks: balance perceived episode loudness, then check the output for clipping and excessive true peaks.
Should I normalise episodes before streaming or use a live compressor?
Prepare files offline when you can inspect episodes and transitions before they enter the playlist. Real-time correction can help with changing material, but it cannot predict the next file and may alter intended dynamics; use it cautiously and listen for pumping.
Why does a normalised episode still sound uneven next to another one?
A similar integrated reading does not guarantee that two programmes have the same speech density, tonal balance or transitions. Listen to the real sequence, including intros, pauses, adverts and music beds, and correct the specific mismatch rather than forcing every peak to match.