If a corpus consists of small speech files which have been recorded without attention to certain details, inconsistent sound quality may compromise its usability.

Homogeneous recordings can be ensured by controlling the average dB values of the speech signal, the average tone, and the speech/noise ratio. A recording studio, or a quiet room lined with sound-absorbing material, would be effective in reducing background noise.

If the amount of material to be recorded during one session is too large, the speaker's tone will dampen. Therefore, the material should also be divided into several smaller sessions of a size with which the speaker feels comfortable.

Recording sessions should be scheduled at a fixed time of the day. A prompting program which displays text at regular intervals is also useful, so the speaker does not need to move his or her face to read during the recording. The speaker should be instructed to maintain the same distance between their mouth and the microphone, unless a head-mounted microphone is used.