A conversation is made up of multiple blocks, with each block representing an audio recording.
In the current version, it is not possible to manually edit the audio content of a single block. This is usually not necessary, as some automatic editing is performed on the audio blocks when exporting the audio. This includes removing the leading and trailing silences, and normalising the volume.
If you still wish to do edit the audio, you can download it and edit it in a digital audio workstation (DAW), then and upload the audio back into the block.