Prepare for the RVR by treating voice writing as three separate layers: a verbatim content stream, a punctuation-and-command stream, and a transcription layer where you resolve sounds-alikes, numbers, and speaker identification from context. Drill each layer in isolation, then combine them in timed dictation with a written self-audit against the source audio.
What the verbatim standard actually requires from a voice writer
Verbatim capture means recording every spoken element exactly as uttered, including false starts, repeated words, hesitations, and corrections, rather than producing a tidied summary of what the speaker apparently meant to say.
A voice writer speaks the record into a mask in real time, so the temptation is to smooth the audio into readable English. Compare the two modes directly: clean reading produces 'I left around three thirty,' while verbatim produces 'I — I left at, uh, around three... no, wait, it was three thirty.' The verbatim version is longer, harder to voice, and it is the one a transcript is judged against.
Train the distinction with a labeled-transcript exercise. Take sixty seconds of recorded speech from a podcast or mock deposition, transcribe it yourself, then mark every hesitation, false start, and repeated word you dropped or reordered. Watching your own omissions accumulate on the page is more persuasive than any rule statement, and it tells you whether your tendency is summarizing, reordering, or silently deleting.
| Feature | Clean reading (summary style) | Verbatim capture (transcript style) |
|---|---|---|
| Hesitations and fillers | Dropped for readability | Captured as spoken: 'uh,' 'um' |
| False starts and corrections | Collapsed into the final wording | Both versions preserved: 'I left at — no, it was — three thirty' |
| Repeated words and stuttered syllables | Condensed | Reproduced: 'I — I — I told him' |
| Purpose in the record | Digest for a reader | Evidence of exactly what was said |
Mask discipline: keeping your record audio clean and contained
Voice writing depends on a sealed mask, controlled speaking volume, and suppressed non-record sounds such as coughs, breaths, and off-microphone comments, because audio quality and containment are part of the professional record.
Mask technique has two goals that pull against each other. The seal must contain your voice so bystanders cannot hear the record being spoken, and the voicing must stay consistent so your speech-to-text output remains accurate. Practice checking the seal before every session: exhale and inhale through the mask and listen for leakage, then run a two-minute read-aloud and note any points where your volume crept up during exciting or fast passages.
Discipline around non-record sounds is a separate habit. A cough, a laugh, or an aside spoken outside the mask either breaks your concentration or leaks into the room. Build a deliberate protocol: pause your content voice, handle the disruption, then resume verbatim without summarizing the gap. Drill this by dictating a passage and deliberately triggering small interruptions — tapping a pen, shifting papers, sipping water — and checking afterward that the record resumed cleanly at the right word.
Separating content voice from punctuation-by-voice and commands
Voice writers layer two streams: the content stream reproduces the speaker, while the command stream handles punctuation, capitalization, formatting, and real-time functions. Keeping these layers distinct is a trainable skill, not an automatic one.
Scenario A: during mock testimony, the witness says, 'The defendant said stop new paragraph and then left.' A plausible mistake is voicing the phrase as if 'new paragraph' were clearly the witness's words, then failing to mark in the transcript that those words were spoken rather than issued as a command. The better handling is to voice the content verbatim, use your established command vocabulary around it, and ensure the transcript shows the witness actually uttered those words. Why it matters: a command spoken by a witness is evidence; a command from you is formatting. The record must distinguish them.
Build the layer distinction deliberately. Write out your command vocabulary — punctuation, paragraph breaks, capitalization, speaker labels, spell-outs — and rehearse saying commands with a noticeably different cadence and placement than content, such as in the natural pauses after clauses rather than mid-word. Then dictate a passage that contains the literal words of your commands and verify in the output that each occurrence was routed to the correct layer. Misrouted commands show up as broken formatting or, worse, as testimony that never made it into the record.
Sounds-alike discrimination: resolving homophones from testimony context
Voice-writing output converts sound to text, so words that sound identical — their/there, ilium/ileum, cite/site — must be resolved using the subject matter and wording around them, not by guessing at the moment of dictation.
Scenario B: a deponent describing hip pain says, 'They found a fracture of the ilium.' A speech recognition output or a hurried voice writer may produce 'ileum,' which is part of the small intestine — an anatomically impossible finding for a hip complaint. The better decision is to catch the mismatch during transcription, weigh the surrounding testimony about the pelvis and orthopedic examination, and select the spelling that the context supports. Why it matters: in a medical-legal record, the two words point to different body systems and different specialists.
Convert this into a recurring drill. Keep a personal list of sounds-alike pairs you personally misroute, drawn from your practice dictations rather than a generic list, and pair each with the contextual cue that disambiguates it — for instance, fracture versus digestive symptom, filing a lawsuit versus a construction location. During transcription, run a targeted pass searching for members of your list and confirm each instance against context. Over time your ear learns to flag the ambiguous sound the moment it is voiced, and your context check becomes a background habit instead of a transcription rescue.
Numbers, names, and addresses: the high-stakes verbatim details
Figures, personal names, street addresses, and spelled-out items are exactly where a record is checked most closely, so voice writers need rehearsed routines for confirming and rendering them accurately.
Numbers demand decisions in real time: spoken figures, fractions, and dates each have conventional transcript renderings, and a speaker who self-corrects — 'nine hundred, no, nineteen hundred dollars' — must be preserved as spoken, not silently corrected. Drill with dictated material rich in figures: record yourself or a peer reading amounts, dates, and case citations at conversation speed, then compare your transcript against the source list item by item.
Names and addresses need a confirming routine. When a speaker states a name or address, the established professional move is to interrupt at an appropriate moment to ask for the spelling, and in practice sessions you should rehearse both sides: recognizing when an unfamiliar name has just passed and voicing a polite, efficient request. In solo drills, practice spelling aloud every proper noun in your dictation passage immediately after it occurs, then check your spellings against the source. This trains the habit of treating every proper noun as unverified until confirmed.
Transcription judgment: speaker identification and record formatting
A voice writer's work is not finished at capture; producing the transcript requires correct speaker labels, clear handling of overlapping speech, and consistent formatting that lets a reader follow the proceedings.
Speaker identification fails quietly in fast exchanges. When two people talk over each other or a witness answers before counsel finishes, the transcript must attribute each utterance correctly, and a voice writer running real-time output may need to fix labels afterward from context and memory of the proceedings. Practice with a two-voice recording: transcribe a rapid back-and-forth exchange, then audit every speaker label against the audio, marking each misattribution and noting what cue in the voices — pitch, speaking pattern, content — you could have used.
Formatting choices also carry meaning. Colloquy between attorneys, the witness's narrative, and nonverbal events such as a document being marked all have conventional transcript treatments, and a reader depends on that consistency to reconstruct what happened. Review practice transcripts specifically for whether nonverbal context is noted where it matters — for example, that an answer referred to a marked exhibit — and whether interruptions are shown as they occurred rather than reordered into tidy turns.
A preparation sequence with a self-audit rubric
Structure preparation in cycles: build each skill layer in isolation, then combine layers under timed dictation, then audit the output against source material using a fixed checklist so every practice session produces a measurable result.
An adaptable sequence runs four weeks per cycle and scales to whatever time you have. Weeks one and two: isolated drills — labeled-transcript verbatim exercises, mask seal checks, command-vocabulary rehearsal, and a sounds-alike list grown from your own errors. Week three: combined practice — short timed dictations that force content and commands simultaneously, followed immediately by transcription and comparison against the source. Week four: full-length simulated sessions with interruptions, numbers, and multiple speakers, then a complete audit and an update to your personal error list. For administrative details such as scheduling and current requirements, rely on NVRA's own site rather than third-party summaries.
Score every simulated session against a rubric rather than a feeling. Use this five-point self-check: every false start and repeated word present in the transcript; no command misrouted as content or vice versa; all numbers and dates matching the source; all proper nouns verified or flagged for confirmation; speaker labels correct throughout. A session that misses two or more points tells you which layer to isolate in the next cycle — that is the purpose of the score: a learning milestone showing where to drill, not a prediction of exam results.
Readiness checks before sitting any assessment: you can complete a ten-minute verbatim dictation with interruptions and score four of five rubric points; your sounds-alike list has stopped growing across the last three sessions; you can voice a spelling request for an unfamiliar name without breaking your content stream; and your command vocabulary is written down and stable rather than improvised.
- Verbatim fidelity: every false start, repetition, and hesitation from the source appears in your transcript
- Layer separation: no command vocabulary appears as testimony and no testimony is swallowed by commands
- Detail accuracy: all figures, dates, and proper nouns match the source or are flagged for confirmation
- Recovery behavior: after any interruption, the record resumes at the correct word with no summarized gap
- Rubric trend: your five-point self-check score rises or holds steady across at least three consecutive sessions
References and further reading
Use these references to explore the concepts and check the latest information from the relevant organizations.