There are two definitions of forensic transcription in use. One use of the term refers to the transcription of any recording that is to be used as evidence in court or even during the process of an investigation. Depending on the recording, the audio quality and intelligibility of the speech may be quite good or extremely difficult to understand.

Many evidentiary recordings are made in a controlled environment with professional grade audio equipment and the permission of those involved. However, other recordings may be made covertly from a distance or in a noisy environment, with poor quality equipment, or by way of inadequate recording methods. These recordings can be quite difficult to decipher and sometimes require the skill and knowledge of a forensic linguist, or more specifically, a forensic phonetician.

Many types of recordings can be considered forensic since anything that is being used as evidence falls under the heading of forensic. Examples include suspect interrogations, witness statements, wire taps, undercover body recordings, general conversations and phone calls, audio journals, and 911 calls. Even transcripts of court hearings and depositions can be considered forensic but are more often referred to as either legal or criminal justice transcripts.

Another aspect of forensic transcription is that recordings must be transcribed verbatim. This means that all speech and other sounds must be transcribed. Even sounds that would normally be perceived as noise could be a contribution to a legal case. Utterances and sounds such as stammering, false starts, you-knows, ums, coughs, ringing phones, car alarms, and so-on are included. Business and medical transcriptionists are not expected to transcribe all of this noise. Verbatim transcription generally takes more time to complete and is a much more tedious task.

Many people, however, use another definition of the term forensic transcription. This second definition explains it as a scientific process used in the analysis and transcription of marginally intelligible speech. Speech may be difficult to understand for any number of reasons. Examples include persons with speech impediments or foreign accents, overlapping speech, distant speakers, poor quality recording devices, concealed microphones, and noise from traffic, crowds, and other sources.

The scientific process of forensic transcription involves professional grade audio equipment and software and is done in a laboratory setting where the forensic examiner can listen without distraction. It can be a very tedious and mentally exhausting process requiring the examiner to listen to sections of audio over and over.

A variety of methods may be employed to decipher marginally intelligible speech. Methods include the altering of the playback speed, amplification, looping, and the use of audio filters to reduce the presence of unwanted frequencies. Also, close attention is paid to syntactic, lexical, phonological, and other aspects of an individuals speech. Finally, a spectrogram, which is a visual representation of a speech waveform, may be used to help identify phonemes (speech sounds) to assist in determining what was said.

This type of forensic transcription is done by forensic audio examiners, forensic linguists, and sometimes speech scientists. It is often done for legal purposes and in such cases it would be considered forensic under either definition. However, not all difficult to decipher recordings are a concern of the legal system. Marginally intelligible recordings may be of importance to historians, anthropologists, journalists, speech pathologists, and many others.