Verbatim Transcription Services

True verbatim captures every word spoken, including filler words ('um', 'you know', 'I, I, I'), false starts, repeated phrases, crosstalk between speakers, audible pauses, and non-verbal sounds where they bear on meaning. Intelligent verbatim, sometimes called clean verbatim, preserves the substance of what each speaker said but removes the filler and smooths false starts into coherent sentences. Two transcripts of the same audio in the two styles will typically differ in word count by twenty to forty percent. We pick the style at intake based on what the transcript has to do next: impeachment cross or witness-credibility work wants true verbatim; arbitration submissions and trial designation read aloud favor intelligent verbatim.

It depends on the document and its destination. Medical depositions used for impeachment cross of a treating physician benefit from true verbatim, because the expert's hesitation pattern under cross is part of the evidence. Medical depositions designated for trial readback to a jury, independent medical examination reports for insurance adjuster review, and life-care planner expert opinions intended for arbitration submission read better in intelligent verbatim, because the substance, not the cadence, is what the audience is evaluating. For raw medical record dictation by physicians (patient notes dictated post-visit), the industry standard is intelligent verbatim, since the goal is a clinically readable chart rather than a record of the dictator's speech habits. We confirm the style at intake and apply it consistently across multi-session matters.

The industry recognizes four main transcript styles. True verbatim captures every word, including filler, false starts, crosstalk, and non-verbal sounds. Intelligent verbatim, also called clean verbatim, preserves substance and removes filler. Edited transcripts go further: redundant sentences are summarized and chronology can be normalized, used in board minutes and executive summary contexts. Phonetic transcripts capture sounds rather than words and are used almost exclusively in forensic linguistics and academic research. For legal work, the relevant choice is between true verbatim and intelligent verbatim. Edited and phonetic transcripts are not court-record use and we do not deliver them as certified transcripts.

The clearest example sits at the top of this page: a hypothetical insurance examination under oath, the same thirty seconds of audio rendered in both styles side by side. True verbatim preserves every filler, every 'I, I, I', every audible pause, and crosstalk markers where one speaker interrupted another. Intelligent verbatim renders the same exchange as a clean question-and-answer. Word count on the true verbatim sample runs roughly thirty percent higher than the intelligent verbatim version. Send a recording with your matter type and we will return a five-minute sample in each style before any commitment, so the choice can be made on actual output rather than on a description of the styles.

A court-admissible verbatim transcript takes six steps. Send the source recording through a secure intake so the file hash, duration, and chain of custody can be captured on arrival. A trained verbatim transcriber produces the first pass in the style chosen at intake, with speaker labels, page-and-line numbering where applicable, exhibit references, and the non-verbal markers documented above. A second reviewer compares the draft against the source recording, normalizes terminology, and flags counsel-judgment items (sealed names, off-record asides, audio gaps). The transcriber signs a certifier's affidavit attesting the transcript is a true and accurate record of the source recording. The chain of custody log is attached to the delivery. The final transcript ships in the format you requested (PDF, DOCX, plain text, or a courtroom-presentation-compatible export) ready for filing.

ChatGPT itself does not transcribe audio. Adjacent tools (OpenAI Whisper, Google Speech-to-Text, Otter, Rev AI) handle audio-to-text and average roughly eighty to ninety percent accuracy on clean recordings. Two problems for legal use. First, AI transcription cannot reliably distinguish true verbatim from intelligent verbatim because it does not understand which speech artifacts matter to the case (a false start preserved for impeachment versus the same false start cleaned for an arbitration submission). Second, AI cannot sign a certifier's affidavit, and a transcript that has to survive a Federal Rules of Evidence 902 challenge needs a human attesting under oath that it is a true and accurate record. We use AI as a first-pass assist on long recordings, then a human transcriber does the verbatim work in the chosen style and signs the certification.