A radio stream can sound uneven because levels drift inside your station chain, or because playback in YouTube changes what a viewer hears. Measure the signal you send first; then check YouTube’s player settings before treating a playback difference as a source or encoder fault.
The practical aim is consistent speech and programme flow without flattening music or other intentional contrasts. Loudness measurement, restrained live correction and peak control address the station’s output; Stable volume and other YouTube enhancements are a separate diagnostic layer.
Separate station drift from playback changes
Start by asking where the difference occurs. Does a quiet song, presenter or advert already sound quieter in the source feed or encoder output, or does the difference appear only when you listen through YouTube? A change heard in the player does not by itself prove that the source level or encoder changed.
This distinction matters because the fixes are different. A low voice compared with the station’s other voices may call for source balancing or modest correction upstream. If the station output is steady but YouTube playback sounds different, changing a compressor or raising the whole mix may make the actual feed less consistent rather than more.
Use a repeatable comparison. Choose a short passage containing speech, music and a transition, and monitor it at the station’s output as well as through YouTube. Keep the playback volume fixed while you compare. If you can, capture or meter the signal before processing, after processing and at the encoder output; note which point first shows a change.
The purpose is not to prove that every listener gets identical playback. Devices, player settings and listening levels vary. It is to identify whether a level shift is present in the signal you control, in the YouTube player, or in both. For a continuous playlist, the choice of playback method and its behaviour over time can also matter; the discussion of vMix and FFmpeg for a continuous playlist stream is useful context for locating where the feed is assembled.
Measure the programme, not just the peaks
A peak meter answers whether a brief instant reaches a high level. It does not tell you whether one song, spoken segment or advert is consistently louder than another across a programme. For that, use a loudness meter that reports integrated loudness. Loudness range helps show how much a programme moves between quieter and louder passages, while true-peak metering estimates peaks between digital samples that a sample-peak meter can miss.
Meter representative material rather than making a decision from a single loud chorus. Include a quieter track, a dense track, spoken introductions, jingles and transitions. A devotional channel might compare a soft bhajan, a full ensemble passage and a spoken dedication. A local news loop might compare a presenter, a report clip and a short music bed. The content varies, but the diagnostic is the same: identify whether the mismatch is programme-wide or limited to a particular item.
EBU R 128, version 5.0 (November 2023), recommends an average programme loudness of -23 LUFS and the use of Loudness Range and Maximum True Peak Level descriptors. That is a broadcast recommendation, not a YouTube-specific target. EBU’s R 128 recommendation and TECH 3401 guidance describe broadcast and radio production contexts; they give you a framework for measuring and controlling your own chain, not a universal number to impose on every platform.
For a station that also feeds radio or another service, document which delivery context each measurement belongs to. EBU’s streaming guidance, R 128 Supplement 2, treats streaming levelling as a distinct context. The useful lesson is to keep the measurement and target tied to the output they are meant to serve. Do not take a broadcast reference and assume YouTube uses it.
Correct live variation with restraint
If your feed is genuinely live, there may be no finished file to normalise before transmission. Real-time loudness correction can help reduce level drift, but it has a limitation: it cannot know whether the next loud passage is an intended musical build or an accidental jump. Audio Engineering Society guidance describes an analysis period of 30 seconds to two minutes as a balance between controlling average loudness and preserving natural dynamics; longer integration tends to sound more natural, with fewer audible artefacts. Those are guidance values, not a preset that suits every format.
Begin upstream by fixing obvious mismatches. Balance presenter microphones, source files and playback devices before asking a processor to compensate for everything. If a batch of old recordings is quiet, gain or normalise those assets individually where possible. That is usually more predictable than making the live processor work harder every time one quiet item appears.
For a live chain, make small changes and listen through representative material. A correction that makes a presenter intelligible can also pull up room noise between phrases. A compressor that restrains a loud segment can make a song feel less open if it works continuously. Keep loudness range in view where preserving dynamics matters, and check transitions as well as steady passages.
A hardware or software audio compressor limiter may be relevant for real-time control and peak limiting, but it is not a substitute for measurement or careful setup. Choose equipment or software only after checking its actual functions, compatibility and operating requirements. If you record separate microphones or sources, keeping separate audio tracks for your live stream can make it easier to find and correct the source of a level mismatch rather than treating the whole programme as one problem.
Limit peaks after applying gain
Raising a quiet signal also raises its peaks. If the programme already approaches the available digital ceiling, added gain can clip the output or leave too little margin for peaks between samples. Apply peak control after the gain stage that creates this risk, then verify the result with a true-peak meter rather than assuming a processor’s output setting tells the whole story.
A limiter can catch brief overshoots; it cannot repair a bad balance between sources or make uneven content naturally consistent. If it is constantly working, first check whether you are applying too much gain, feeding it an already compressed source, or asking one setting to handle sharply different material. Listen for distortion, pumping and a sudden loss of impact, especially on drums, cymbals and sung peaks.
Treat loudness and maximum true peak as different checks. EBU R 128 uses both programme loudness and a maximum true-peak descriptor for a reason: a signal can have a sensible average level and still contain a troublesome peak. Conversely, keeping peaks low does not guarantee that a quiet programme will sound balanced beside the next item. Meter both, then audition the transition.
Check Stable volume and other YouTube enhancements
YouTube documents viewer-side audio enhancements that can include automatic adjustments to volume levels and sound mixing. Stable volume is intended to reduce differences between quiet and loud parts; Voice boost can reduce background sound and highlight speech. These features are not applied to every video, so a change in one player session is not evidence that all viewers receive the same treatment.
YouTube Help says Stable volume is on by default when watching videos where the feature is available. It is not available on every video, is turned off for YouTube Music and official music videos, and can be toggled in player settings. When Voice boost is on, Stable volume turns off. Check YouTube’s current Stable volume and Voice boost controls while reproducing the issue, rather than relying on a remembered setting.
YouTube also says some enhancements may be managed for future uploads in YouTube Studio. The setting does not affect ads or videos already uploaded, and viewer accessibility settings can take precedence in some cases. Consult the current YouTube Help page on audio enhancements for the available controls and their limits. Do not assume that a channel-side setting will correct a live signal, or that every enhancement is available for every piece of content.
When a viewer reports a shift, ask which device and player they used, whether Stable volume or Voice boost was enabled, and whether they heard the difference in the live player or a replay. These questions do not establish a universal playback path; they help separate a repeatable station fault from a setting or device-specific observation.
Compare encoder output with what viewers hear
Make a simple comparison record. At the station, note the loudness and true peak at the relevant output point. At the viewer end, replay the same passage in YouTube and match listening volume as closely as practical. Do not compare a loud pair of headphones with a quiet phone speaker and call the difference a change in YouTube’s level; device playback and listening conditions confound the comparison.
Use the sequence below when the report is intermittent:
| Check | What to compare | What it suggests |
|---|---|---|
| Source material | The same passage before station processing | A mismatch here points towards the file, microphone or source balance. |
| Processed station output | Loudness, loudness range and true peak after correction | A new mismatch here points towards processing or gain staging. |
| Encoder output | The same passage leaving the encoder | If it differs from the processed feed, inspect routing and encoder configuration. |
| YouTube player | Same passage with player enhancements noted | A difference confined to playback calls for a player-side check before station changes. |
A meter reading cannot tell you whether speech remains clear or music still feels natural. Listen to speech, music, jingles and adverts at matched playback volume. If the station output meter is steady and the difference happens only in YouTube, test the player’s available audio settings before altering your station chain. If both outputs show a shift, investigate the earliest point where the measurement changes.
This is also a good time to verify that the stream itself is stable rather than confusing a drop or interruption with a volume problem. The practical checks in monitoring a YouTube live stream for dropped frames and disconnects can help distinguish a transport interruption from a change in audio level.
Do not match unlike processing targets
A station may have a broadcast output, a web stream, a YouTube live output and archived programme files. They need not all use the same target or processing. EBU R 128’s -23 LUFS recommendation is useful as a broadcast reference where appropriate; it is not documentation of a YouTube mastering target. EBU’s streaming guidance notes that streaming platforms typically use a higher target loudness level than broadcast standards, which is another reason not to force one number onto every delivery path.
Set a target only after deciding what output it serves and how it will be measured. Keep a record of the meter, measurement point, target and peak ceiling for each chain. If you operate a FM or DAB service alongside YouTube, the web output may need a separate delivery choice. If your YouTube channel mainly carries music, preserving dynamic contrast may matter more than making every passage equally loud. A news loop with frequent spoken links may have a different editorial balance.
Do not try to match a guessed YouTube normalisation value by driving your own output harder. YouTube’s documented playback enhancements do not establish that it applies your station’s mastering target, and there is no universal current LUFS target in the cited help guidance. Keep the station-side signal measured and controlled, then treat the player as a separate check.
A repeatable check before changing settings
When a report arrives, capture enough detail to reproduce it: the programme item, timestamp, station output point, encoder output, player, device and enhancement settings. Start by checking source levels and transitions, then measure integrated loudness, loudness range where relevant and true peaks through the station chain. If the change first appears after your processing, make one restrained adjustment and repeat the same passage.
For live correction, avoid changing several controls at once. A modest adjustment followed by a listening check tells you more than a broad gain increase plus aggressive compression and limiting. Confirm that the limiter is not constantly reducing peaks, that voices remain clear, and that quieter passages have not become noisy. Keep intentional contrast between songs or programme segments rather than treating all variation as an error.
If you cannot reproduce the difference in your own output, test YouTube playback with Stable volume and Voice boost states noted. Compare at matched playback level and use a second device only as an additional observation, not as proof of the station’s target. Repeat the check with a later item if the first passage is unusual. Record what changed so a night-time report can be acted on without guessing.
For an always-on channel built from a prepared file, the operating burden may be less about a live desk and more about keeping the broadcast going while your computer is off. StreamNeo removes that specific need to leave a personal computer running by turning an uploaded video into a continuing YouTube live stream; it does not replace measuring or preparing the audio file, and the service is for YouTube only.
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
Does YouTube change the volume of every radio stream?
No. YouTube documents Stable volume and other audio enhancements, but they are not available or applied in every case. A perceived difference in the player does not prove the station’s source or encoder level changed.
Should I set my stream to -23 LUFS?
EBU R 128 recommends -23 LUFS average programme loudness as a broadcast reference, not as a universal YouTube target. Choose a measurement target for the delivery chain it is intended to serve, and check current official guidance for your context.
Will a limiter make my radio stream consistent?
A limiter helps control peaks after gain is applied, but it does not balance unlike sources or fix every loudness mismatch. Measure programme loudness as well as true peaks, and listen for pumping or loss of intentional dynamics.
What should I ask a viewer who reports a volume change?
Ask which passage, device and player they heard, and whether Stable volume or Voice boost was enabled. Compare that passage with your encoder output at a matched listening level before changing station processing.