In a two-person conversation, one speaker may be quiet and the other much louder. Separate microphone tracks allow later adjustment without necessarily changing both voices together. However, a recorder may save only a combined mix despite having two microphones connected. Check the actual file result as well as the plugs.
Assign each person to a channel
Label inputs with names or roles. Have only person A speak, followed by only person B. Watch the meters and record a short test. Open it in the editor: the intended isolated tracks must actually exist. Some devices store several channels in one multichannel file; others create individual files.
Set input gain separately for each voice. Different control positions are normal. The aim is not matching knob positions, but clean recordings with suitable headroom. Also listen to brief simultaneous speech, because interruptions occur in real conversations.
Use distance and orientation to reduce spill
Each microphone should be close to its own voice and, where possible, less sensitive towards the other voice. The precise arrangement depends on its pickup pattern. Do not simply put both microphones together in the middle of the table and expect complete isolation.
Some spill usually remains in a shared room. With both tracks open, the second, more distant recording of a voice can alter its sound. Increasing the distance between people helps only while conversation remains natural and the microphones stay sufficiently close. Test the seating arrangement with eye contact, not merely with both heads facing forwards.
Move naturally between voices in editing
Work on the tracks separately initially, then listen together. Reducing the currently quiet channel can lower room noise, but must not remove interjections or the beginnings of breaths. Aggressive automatic gates can swallow short opening words. Manual transitions or restrained automation are often easier to judge.
For an ordinary conversational podcast, both voices can remain centred in the stereo image. Recording them on left and right is technical routing, not a requirement for publication. Check the finished mix in mono and through one loudspeaker too. Every speaker should remain fully intelligible.
Keep the untouched isolated tracks and a clear channel note. A combined export is useful for publishing but does not replace separate originals for later corrections. After a break, check again that nobody has muted the wrong input or switched recording to a summed mix.
An example with overlapping speech
Speaker A explains a procedure while speaker B briefly adds “exactly”. Track A contains the main voice clearly and the interjection quietly; track B contains the reverse. Normalising both tracks to the same level does not solve the scene, because it also raises their distant voices. Listen separately first, choose the close recording for each statement and retain the interjection where it belongs in the conversation.
When a word remains audible on both tracks, a small timing difference can alter its tone. Do not hastily move an entire track, because its own main voice may then lose synchronisation. Instead assess levels and transitions in the particular passage. Carefully reducing the less important channel may suffice. For future recordings, improved microphone placement is often a more dependable remedy than elaborate phase processing afterwards.