The introduction sounds confident, the first guest sounds distant, and the music between sections makes the listener reach for the volume control again. On the editing screen, nothing looks obviously wrong. Each section has a healthy-looking waveform. The problem is that similar peaks do not necessarily create a consistent listening experience. A production can be individually loud in all the right places and still feel uneven as a whole.
For a small podcast, the practical aim is simple: let people follow the conversation without continually compensating for the edit. That aim involves the relationship between voices, transitions and the complete programme. It cannot be reduced to enlarging every waveform or applying the same effect separately to every clip. Measurement helps, but only when the measurement answers the question you are asking.
This is a workflow explanation rather than a listening test of a particular product. The examples are hypothetical. They illustrate decisions an editor can make before choosing a final delivery level, which should follow the requirements of the actual destination. A setting suitable for one broadcast specification should not quietly become a universal target for every podcast service.
A peak is an event; a programme is an experience
The EBU’s overview of loudness explains the move from relying on peak meters alone to loudness measurement and normalisation. Its R 128 recommendation belongs to a documented broadcast framework, with additional guidance for different applications. The relevant lesson for a creator is the distinction between measures, not an instruction to copy a broadcast number into every project.
In an imagined interview, one voice contains an occasional sharp sound while another remains relatively steady. The highest point of each recording does not describe everything between those points. Matching only the most extreme moment can leave the conversational passages feeling different. The listener experiences the sentences, not just the single highest sample represented on the screen.
That is why a large waveform can be reassuring without being sufficient. It shows a representation of amplitude over time, but it does not make every editorial question visible. You still need to hear how a guest enters after the presenter and how the programme changes when music appears. The relationship is the issue.
Before adjusting anything, identify where the inconsistency occurs. Is one speaker difficult to follow throughout, does a single passage drop away, or does a transition create the impression of a sudden change? These observations point to different decisions. Applying a final process before understanding the pattern can make it harder to tell what you are trying to correct.
Separate repair, balance and delivery
An audio project can contain problems that no overall loudness target explains. An unclear word, an awkward edit and a distracting background sound are different issues from the final programme level. Treating them as one problem encourages a chain of effects without a clear purpose. Start by naming what the listener needs to understand.
The Audacity manual for Loudness Normalization distinguishes the effect from peak normalisation and describes setting a LUFS target. It also notes that applicable targets vary by use and presents the effect as normally a final step before exporting production audio. Those details support a useful separation between preparing a mix and meeting a delivery requirement.
In a hypothetical workflow, the editor first resolves intelligibility and obvious edit problems, then compares the speakers within the conversation, and finally evaluates the finished programme against its destination’s specification. These are not compulsory button presses in a fixed recipe. They are different questions that should remain distinguishable when you decide what to change.
If a guest recording is intrinsically poor, describe its limitations honestly rather than promising that normalisation will reconstruct missing information. Increasing level is not the same as recovering an absent word. You may need another source recording, clarification or an editorial decision about whether that passage can be used responsibly.
This matters when time is limited. A project can absorb hours of adjustment because the editor keeps using a level control to solve a problem of clarity. Returning to the original question can reveal that the best action is a different edit or a more suitable recording, rather than another round of processing.
Review the boundaries, not only the clips
An interview excerpt can sound comfortable on its own and still feel abrupt when placed after a different voice. Listen across the boundary. Include enough material before and after the edit to understand the change as a listener will hear it. A selection that begins exactly at the clip can hide the relationship you need to assess.
Consider an invented episode with a presenter introduction, a remote guest and a short musical transition. Reviewing each item separately answers whether it is usable in isolation. Reviewing the sequence answers whether the listener can remain with the story. The second question is particularly important when the episode contains material recorded in different settings.
Keep the comparison conditions reasonably consistent while making judgements. If you keep changing the playback level between clips, it becomes harder to evaluate the relationship you are editing. This is a practical comparison principle, not a prescribed listening volume or a claim that one device reproduces every audience experience.
Write a short note about the observed issue before moving controls. For example, the guest entrance feels much quieter than the preceding sentence. That note is more actionable than make the podcast professional. It also gives you a way to judge whether the change solved the problem instead of merely making the waveform look different.
Once you have changed a transition, return to the surrounding passage. An isolated improvement can disturb the broader sequence. The aim is not to make every sound equally prominent. It is to preserve the intended relationship between information, emphasis and atmosphere so that the listener does not have to repair the edit with the volume control.
Music should have a role you can describe
Music can mark a chapter, create a mood or separate parts of a programme. Those are editorial choices. If it overwhelms the first sentence after a transition, the listener may miss the information the transition was meant to introduce. The question is whether the sound supports the structure, not whether the music file is impressive by itself.
In a hypothetical programme, a short musical passage ends under the presenter’s return. Listen to the words at that intersection. If they are difficult to follow, reconsider the relationship or the edit. You do not need a claim about a universal ideal music-to-speech number to recognise that the audience is being asked to process two competing elements.
Silence can also be intentional. A pause after a difficult answer may serve the conversation, while an unexplained gap between sections may feel like a playback problem. Do not remove every quiet moment merely to make a display look continuous. Decide what the moment communicates in the episode and edit accordingly.
The same principle applies to emphasis within a voice. A speaker’s changing delivery can carry meaning. Flattening every difference may create a more uniform signal while weakening the human rhythm of the exchange. Any processing should be judged in relation to intelligibility and expression, rather than to the goal of erasing every variation.
When several people edit, share those intentions. A producer may leave a pause deliberately, while a later editor assumes it is an oversight. Notes about purpose prevent a technical cleanup from silently changing the storytelling. A meter cannot explain why the pause was there, so that information needs to come from the project.
The final specification needs a named destination
Before choosing an export target, consult the current requirements of the service or recipient that will use the file. Record which document you are following and what it asks for. A number remembered from a tutorial is not a specification for every destination. It may describe a different kind of programme or distribution route.
The EBU’s broadcast recommendation, for example, is not evidence that every independent podcast must use that same target. Nor should an editor assume that a platform’s playback processing makes delivery choices irrelevant. The appropriate decision depends on the documented destination and the production’s own workflow, not on a general slogan about what the internet prefers.
In an invented client project, the recipient supplies an audio delivery sheet. The editor checks that the final file corresponds to that sheet and notes any ambiguity before delivery. If no requirement is supplied, the next step is to establish a suitable basis for the project, rather than declaring an arbitrary preset an industry-wide rule.
Keep measurement conditions understandable too. Audacity’s documentation describes options affecting mono and stereo treatment. If two readings differ, inspect the relevant settings and material before concluding that one tool is wrong. A comparison is useful only when you know what is being compared and how it is measured.
This article does not prescribe a particular set of switches for every production. It recommends retaining enough information to reproduce your own result. A clear record of the selected target and relevant settings is more useful for the next episode than an unexplained file called final-louder.
Listen to the export you intend to publish
The editing session is not the published asset. After exporting, open the actual file and check the passages that motivated the work: the introduction, guest entrances, transitions and ending. This verifies a result within your workflow. It should not be described as proof that every listening device or environment will reproduce it identically.
Confirm that the intended version was exported and that no section was inadvertently omitted. A technically consistent file can still be the wrong edit. File naming and a short delivery record help separate content approval from audio measurement, particularly when several revisions exist. Both questions matter, and neither automatically settles the other.
A second listening context can reveal an issue you want to investigate, but do not turn one casual comparison into a published product test. Note what you observed and return to the relevant passage. The aim is to catch problems in the programme, not to claim a universal ranking of headphones or speakers you have not systematically evaluated.
If the episode changes after the final check, identify whether the change affects the audio you assessed. Adding a new introduction or replacing a transition creates a new sequence. Revisit the relevant checks for that change rather than assuming that the previous export’s approval transfers to every later revision.
For recurring work, keep the process small enough to repeat. A consistent review of important boundaries, a documented destination and a check of the real export can be more effective than an elaborate chain nobody understands. The process should make decisions visible, not hide them behind a preset whose purpose has been forgotten.
An even podcast does not demand that every second sound the same. It lets the listener follow meaningful changes without unnecessary effort created by the production. Peaks, loudness measurements and listening each contribute different evidence. The finished programme improves when the editor knows which question each one can answer and keeps the conversation at the centre of the work.