The loudest part of a conversation is rarely the line chosen for a soundcheck. A quiet introduction may be followed by a burst of laughter or a sharply emphasized word. Digital recording provides a ceiling above which the incoming signal cannot be captured accurately. Leave space for the ordinary surprises of speech, and use monitoring to distinguish a quiet signal from a quiet pair of headphones.
Understand the meter
A digital peak meter shows how close the signal comes to the recording ceiling. Reaching that ceiling can produce flattened, harsh peaks that are difficult to repair.
It is worth deciding what a satisfactory result would look like before proceeding. Watch the meter during speech rather than judging only headphone volume. A clear expectation makes it easier to recognize when a small alteration has done enough and another is unnecessary. Try not to confuse a conspicuous change with an actual improvement. A clean signal leaves margin for a laugh or sudden emphasis. If the outcome is difficult to assess, repeat the exercise with a simpler example and keep the two versions separate. The comparison is useful even when neither version is ready for final use.
Set levels from realistic speech
A speaker's voice during a brief test can be much quieter than their voice during an enthusiastic conversation. Test with natural dynamics rather than repeated whispered words.
An informative comparison does not have to be complicated. Ask both speakers to tell a short story in their normal conversation tone. Keep the surrounding conditions as similar as practical and preserve the original material when you can. Allow a little time before judging the result, because a recent change sometimes seems more successful simply because it is different. Observe the loudest peaks and adjust the input to leave room. If the improvement is uncertain, a second representative example may reveal whether the issue is really addressed. This is also a good opportunity to discover which details deserve to be retained in your notes.
Do not confuse monitoring with gain
A headphone volume control can make the sound louder in your ears without changing the recorded signal at all. An input gain control acts on the signal before capture.
There is little benefit in spending a long time perfecting an example that cannot be used later. Make a short recording while varying only the headphone level. Think about the practical circumstances in which the result needs to work, and allow for their normal variation. The saved file should have the same peak level in both tests. A second attempt can be especially useful when the outcome was unexpectedly good or bad. It provides a chance to check whether the explanation fits the result. Do not be afraid to leave a less important detail alone when it is already satisfactory.

Further reading in the journal: Where to Place a Microphone at Home.
Check each input separately
A guest and host rarely speak at identical levels. Independent tracks may need different gain settings to preserve each voice cleanly.
Have each speaker talk alone, then run a short conversation together. Try not to judge the success of the decision solely by how easy it is to carry out. Some seemingly convenient choices require substantial reworking later, while a little preparation may eliminate a recurring difficulty. Consider the next person or the next stage that will depend on this result. Verify that both tracks retain room for louder moments. A useful way to decide is to imagine having to return to the work after several weeks. Leave enough context to make that return straightforward.
Allow a conservative margin
A quiet clean track can usually be raised later, provided background noise is acceptable. An overloaded converter cannot recover the original waveform merely by lowering the playback level.
Avoid recording unusually close to the digital ceiling; leave several decibels for unexpected peaks. One advantage of doing this carefully is that a later decision can be based on more than the impression left by the most recent attempt. Keep a concise record that identifies what was changed and what was observed. Compare the strongest parts of the test and verify they sound natural. It is particularly important not to attribute an improvement to the wrong cause when several things happened together. Try a controlled comparison if the explanation is uncertain; a short sensible check is usually less work than correcting a mistaken assumption throughout the project.
Listen as well as look
A meter cannot tell you whether a cable crackles or a microphone is too close to a source of breath. Headphones give context that the meter does not display.
Try to separate the parts that require precise judgment from those that can reasonably vary. Monitor the test while gently checking normal speaker movements. This keeps attention available for the places where a small mistake has a significant consequence. Mark any noise that appears independently of the voice level. If a detail does not affect the intended result but takes an unreasonable amount of time to refine, reconsider whether it deserves further attention. A reliable outcome does not require every intermediate choice to be identical. It does require a clear reason for the decisions that matter.
For a related part of the process, see The Basics of Equalizing a Podcast Voice.
Keep original recordings
A processed or normalized copy may conceal how the recording was made. The original offers a reference if unfamiliar distortion appears later in the edit.
Save the untouched tracks before applying corrective changes. Now compare the proposed solution with the actual needs of the finished work. A technically interesting alteration is not necessarily an editorial or practical improvement. Check that the original remains readable and separate from the working session. Evaluate the complete result, including the portions that were not changed, to make sure the choice did not create an unexpected difference elsewhere. If everything behaves as intended, stop adjusting it. The ability to decide that something is finished is an important part of a dependable independent workflow.
Repeat the test after changes
Moving a microphone or swapping an input device can change the signal in unexpected ways. Recheck gain before proceeding with the actual conversation.
Run a brief test whenever speaker position or equipment changes. Review the result as a whole instead of concentrating only on the most recent adjustment. This is a useful moment to distinguish essential checks from personal preferences that have little effect on the finished material. Both the peak readings and audible voice quality should remain acceptable. If an issue remains, describe it precisely enough to guide a later attempt. Keeping a concise record of the unresolved problem can be more helpful than continuing to make small changes without a clear expectation of what will improve.
Putting the method into practice
A level setting is not permanent. Check it when the guest, microphone position or input arrangement changes, and keep a short recording of representative speech before starting the programme.
