Court Reporting Setups by Job Type: What Each Method Needs, and What the Rules Allow

Every job starts with two questions: what will the record be, and who will sign it? California’s Court Reporters Board answers the first for two of the three methods: “The stenographic/voice notes are the official record.”¹ For digital reporting, the recording is the official record where the rules make it so, as in Florida’s Eighteenth Circuit, and in federal court a court-approved transcriber certifies the transcript made from it.²³ The gear follows from those answers, and so do the rules.

This guide is for stenographic reporters, voice writers, digital reporters and transcribers, and the agencies that book them. It is organized by job, and within each job it says what changes for each method. It closes with the rules and credentials that decide which method may take a job at all.

Each method has its own professional body and credentials: NCRA’s RPR for “Current or aspiring stenographic reporters,”⁴ NVRA’s voice-writing exams,⁵ and AAERT’s exams for digital capture, depositions, and transcription.⁶ Where the bodies disagree, we attribute each position to the body that holds it. Products appear only as examples of what their makers’ own documentation says, not as recommendations or rankings, and we quote no prices.

One thing does not change with the method. Federal law puts the certificate on “The reporter or other individual designated to produce the record,”⁷ and NCRA’s 2026 position statement holds that AI and speech recognition “cannot replace a trained, certified, impartial human court reporter and/or captioner in the creation, preservation, and certification of the official record.”⁸ Our reading: software and recordings support the record. The reporter or the transcriber makes it and signs it.

To turn this guide into a job sheet, use the companion setup picker: pick your method and job, and print a checklist where every line links back to the section that sources it.

At a glance

This matrix is DepoStack’s synthesis of the sources cited, not an industry standard. Find the job, read across to your method, and follow the link. Where a rule may bar or limit a method, the cell points to the rules table, current as of October 13, 2026.

“Check your jurisdiction” means the method question comes before the gear question. California takes depositions stenographically “Unless the parties agree or the court orders otherwise”;¹⁴ Georgia, unless the court orders otherwise;⁴⁴ and Massachusetts, unless the parties stipulate.⁴⁵

What the record is

Steno and voice: the notes are the record

California’s Court Reporters Board says the Board “looks to the transcript and the stenographic/voice notes, not an audio file,” and “do NOT rely upon backup audio media for transcript production.”¹ Its definition of backup audio includes “the audio synchronization tool built into computer-aided translation (CAT) software”; “playback of the backup audio media in lieu of readback is not allowed”; and discretionary backup audio “is the personal property of the CSR.”⁴⁶ NCRA’s guidelines agree: read back from the notes; the backup audio is the reporter’s personal property; and “A reporting firm/agency may not require that a reporter produce the backup audio media (unless ordered to do so by a court).”⁴⁷ The Board’s pointers are guidance, not regulations.¹

A voice writer’s notes are the voice writer’s own. California defines voice writing as a record made with “a closed microphone voice dictation silencer, steno mask, or similar device using oral shorthand and voice notes made by a certified shorthand reporter.”⁴⁸ The Board explains that voice notes “are not a live or full recording of the actual proceeding.”⁴⁹ Federal policy describes stenomask reporting with a “special face-mask microphone” and says “Certified realtime stenomask reporters may provide realtime services.”³

One vendor reads readback differently. ProCAT’s Winner Voice Edition page says the reporter “may play back the room recording track or voice track in response to requests to read b[ack],”⁵⁰ which sits against California’s and NCRA’s bar on playback in lieu of readback.⁴⁶⁴⁷ For backup audio as a checking aid during editing, see The Art of the Transcript.

Where audio becomes the record

In federal court, a reporter’s convenience backup recordings “are the personal property of the court reporter,” but a backup filed with the clerk as the record of an arraignment, plea, or sentencing is a public record that “must be accompanied by a certification.”³ Court-provided sound recording “may not be used to back up court reporters, who are required by law to furnish their own equipment,” nor to give the court a backup when a reporter is recording, “as there can only be one official original record of the proceedings.”³

Minnesota goes the other way. Its Judicial Council policy, in the version effective July 1, 2024, requires an audio backup of every district-court proceeding taken stenographically, “on a court owned and approved audio recording system,” which can include court-licensed video-conferencing software; audio recorded on the steno writer does not satisfy it.⁵¹ A newer revision likely exists; we read the 2024 text.⁵¹

Our reading: find out which rule applies before you decide where your audio goes and who may have it.

Digital: the recording is the record

Where digital recording is the record, it is monitored and logged as it is made. A federal electronic court recorder must “Maintain a log of the proceeding to be retained as an aid to the transcription of the record,” and “court-approved transcriber(s),” not the system, certify “a correct transcript from the official electronic sound recording.”³ The clerk keeps the original notes or records “for not less than ten years.”⁷

Georgia’s 2026 superior-court rule sends the recording “to a certified court reporter who is authorized by the Board of Court Reporting to create transcripts,” and when that reporter was not present, the certificate says the transcript was “generated from a digital recording made outside my presence.” The rule “does not apply to depositions.”⁵² AAERT’s deposition-reporter exam defines the job as a quality audio recording, administering oaths, “time-synced notes,” live playback, and “marking and managing exhibits.”⁵³

AI and speech recognition

NCRA’s 2026 guidance, addressed to stenographic reporters and captioners, says “AI may be used only to support the stenographic reporter’s work and not to generate the record.”⁵⁴ AAERT frames its view around oversight rather than method: “The true gold standard in court reporting has never been the tool. It has always been professional oversight.”⁵⁵ Texas’s 2026 court-recording study reports “near-universal agreement” among respondents “that AI alone is insufficient to produce a legally sufficient record.”⁵⁶ Whether a Texas trial court abused its discretion in striking a non-stenographically recorded, AI-generated deposition transcript is pending before the Supreme Court of Texas, argued October 6, 2026 and undecided as of October 13, 2026 (see In flux).⁵⁷ The longer treatment is in The Art of the Transcript.

Core kit

Equipment here is described by category and specification. A named product is an example of what its maker’s documentation says, not a recommendation; check the current documentation for your own model.

Core kit: steno

Writer and connection. Stenograph documents its NexGen writer as “Realtime ready through USB, virtual serial port, Bluetooth, and Wi-Fi.”⁹ The links differ when a cable comes out. The NexGen guide says that if you unplug USB for a bench conference and reconnect, “the USB realtime connection will feed the strokes written during the bench conference,” and tells users of other CAT programs to check with their CAT vendor;⁹ a virtual serial port connection over USB does not survive unplugging.⁴¹ Stenograph also says its writers cannot establish Wi-Fi communications with CATalyst on a “wireless client isolation” network, which stops devices on the same network from talking to each other.⁹ An NCRA Technology Committee member, relaying a vendor trainer’s tip for one writer, advises pairing Bluetooth even if you use USB, “so you have options.”¹⁰

Storage. Stenograph’s Luminex II will not operate without its Backup SD card, which “is not intended for archiving steno notes” and overwrites its oldest files when full; Stenograph “recommends carrying an extra SD card,” and a laptop’s built-in slot “cannot read the Backup SD card.”⁴¹ Its older Wave guide describes only internal Memory of about 12,000 strokes, “approximately 50 pages.”⁵⁸ Stenovations’ Zenith “creates backup files on a removable SD card in the writer itself.”⁵⁹ ProCAT advertises “four physical drives” for its Xpression.⁶⁰

Power and audio. Battery figures are in the battery table. In Stenograph’s FailSafe mode, the Luminex II turns off its screen, Bluetooth, and Wi-Fi but keeps writing notes and audio to both SD cards, and keeps realtime going over USB.⁴¹ Some writers draw power from the laptop: on the Zenith, “The USB cable provides power to the writer.”⁵⁹ For writer audio, Stenograph says “Record Audio During Job” “must be set to On,” and suggests starting gain “between 75% to 85%” on NexGen and Luminex writers.⁶¹

CAT laptop. Advantage recommends a business-class system for Eclipse (“Intel i5 or better, 3.0 GHz or faster,” “16 GB minimum” of RAM, an SSD) with multiple USB ports, notes “Older writers may require a serial adapter,” and tells MacBook users “M1/M2 Macs are not compatible.”⁶² An NCRA Technology Committee column says “Chromebooks are not compatible with any CAT software.”⁶³ CAT basics are in What Is CAT Software?.

Core kit: voice

A voice writer’s kit is a chain: mask, audio interface, recognition engine, voice CAT, and computer. The links have to agree.

Mask. Talk Technologies, which sells through talktech.com, documents a two-plug Stenomask whose plugs “can be used in conjunction with two separate devices simultaneously i.e; a digital recorder and laptop.”¹³ Martel documents a mask with “two independent microphone outputs, one DIGITAL and one 3.5mm.”⁶⁴ Talk Technologies’ quick-start guide says “If air escapes, sound escapes!” and sets the level at “Quieter than a normal speaking voice but above a whisper.”⁶⁵ Its manual describes “lagging”: waiting several words before repeating, to cut false starts.⁶⁶

Audio interface. Talk Technologies’ FAQ says a Stenomask with the Andrea USB Adapter “works extremely well” with engines such as Dragon, that a standard sound card “will not provide as good recognition as with a USB adapter,” and that “Cable issues are extremely common”; it gives no Windows 11 compatibility statement.⁶⁷ Nuance says Dragon 16 needs “An audio input device capable of 16-bit recording”⁶⁸ and that a USB microphone should plug in directly, “not a USB extension or hub.”⁶⁹ To keep the room on its own track you need two sound devices: Stenograph says a USB mask “has a built-in sound card,” and the wrong room device leaves you with “two recordings of your voice.”¹²

Engine and voice CAT. Nuance presents Dragon Professional v16 as “optimized for Windows 11.”⁷⁰ It stopped selling the English and German v15 editions after February 27, 2023, and ended v14 support on September 1, 2023.⁷¹ Stenograph’s CATalyst VP requires VP version 21 or later with Dragon Professional 16 on the same Windows 11 computer, “16 GB of RAM or more,” and no ARM (Snapdragon) processor; it says to turn off Dragon’s automatic microphone level and to dictate in Dictation mode, because otherwise “Dragon may interpret a word, like ‘copy’, as a Windows command.”¹² Advantage’s EclipseVox supports Dragon and Speechmatics, “with the option to process speech in the cloud or locally,” where local processing requires Advantage hardware.⁷² For voice writing, Advantage recommends a Ryzen 7 or i7 at 3.5 GHz or faster and “16 GB minimum, 32 GB recommended,” and calls Dragon 16 “essential for realtime performance”;⁶² its EclipseVox page, though, lists Dragon “version 15 or later” as required.⁷² ProCAT’s Winner Voice Edition records “the court reporter’s voice on a separate track” and lets the user switch between voice and steno during a realtime session.⁵⁰

Quiet. NVRA’s exam rules say “Participants may not be audible within two feet” and that built-in computer or recorder microphones must be disabled: “Turning the microphone to the lowest volume will not suffice.” During the test, a backup may record only the tester’s voice, through a handheld recorder attached to the mask.⁷³ Those are test rules, not job rules. Our reading: the two-foot line is still a useful quiet check.

What the engine vendors do not say yet. We found no end-of-sale or end-of-support date for desktop Dragon v16 on Nuance’s v16 pages.⁷¹⁷⁰ Advantage’s Boost terms say “Boost requires an active internet connection during use,” but its published terms do not address speech audio sent for cloud processing.⁷⁴ If either matters to a client, ask the vendor in writing (our suggestion).

Core kit: digital

A digital reporter’s kit is a recording system: microphones, an interface or recorder that keeps each on its own channel, software with log notes, headphones, and an independent backup. Microphones, channels, levels, formats, and monitoring are covered under Audio capture.

Channels. For depositions, AAERT’s blog says “you’ll need enough channels to capture each speaking party on their own microphone, plus a table mic or two for backup.”¹⁵ BlueLedge, a training provider, recommends “a one-to-one ratio of microphones to participants,” using “phantom-powered (48V) lavalieres or hand-held microphones.”¹⁶ Two vendor examples of channel counts: SoniClear documents “up to 12 independent microphone channels,”⁷⁵ and For The Record documents “up to 16 channels” for its FTR Reporter application.⁷⁶

Log notes. Federal policy requires the software to “time stamp the log notes when a court recorder begins the log note rather than at the completion of the note.”³ NCSC suggests log entries for case data, each witness, objections and rulings, speaker changes, exhibits, and going on and off the record, plus an appearance sheet with “the channel designation and seating arrangement for all parties.”³⁴ Federal policy also requires export “in a non-proprietary format” that transcribers can use “without requiring the transcribers to purchase proprietary hardware or software.”³

Backup. AAERT’s blog says “You’ll want at least one backup recording in addition to your main recording,” and suggests three.¹⁵ BlueLedge says “every reporter, no matter the methodology of capture, needs a backup recorder,” outside the main audio chain and battery powered, and recommends “at least two forms of backup recording.”¹⁶ Federal fallbacks for a failed system run from “a spare computer with the digital recording system pre-loaded” to “a backup analog recorder and hand-written log notes.”³

What the exam expects. AAERT’s CDR blueprint covers audio interfaces, power sources, confidence monitoring, channel assignment, and “live and dead rooms.”⁵³ AAERT’s Best Practices Guide is still the 2019 edition; a 2026 edition was announced but had not appeared by October 13, 2026.⁷⁷ We did not review the paid guide.

Core kit checklist. DepoStack’s synthesis of the sources cited, not an industry standard.

• Steno: the writer, charged, and its USB cable even if you use Bluetooth.¹⁰
• Steno: a spare SD card, if your writer uses one.⁴¹
• Steno: writer audio on, gain at the maker’s starting level.⁶¹
• Steno and voice: a laptop at or above your CAT vendor’s current spec.⁶²¹²
• Voice: a mask with two outputs, and its USB adapter or a USB mask.¹³⁶⁷
• Voice: the engine version your voice CAT requires, and a second sound device for the room.¹²
• Digital: a mic per speaking party, plus a table mic or two.¹⁵
• Digital: recording software that time-stamps log notes when you start them.³
• Digital: headphones that are not noise-cancelling.¹⁶
• Every method: an independent, battery-powered backup recorder.¹⁶

In-person depositions

Before the gear, the notice. Under the federal rules, the noticing party “must state in the notice the method for recording the testimony,” and “Unless the court orders otherwise, testimony may be recorded by audio, audiovisual, or stenographic means”; any party may add a method on prior notice.²⁸ States differ: see Rules by jurisdiction.

In person: steno

The in-person steno setup is the writer, the CAT laptop, the writer’s audio, and, on realtime jobs, a path to counsel’s screens. Stenograph says “you can use a Wi-Fi router to output CaseViewNet to recipients in the same room.”¹¹ Advantage’s Bridge Mobile serves “Unlimited viewers (local + remote)” with SSL encryption,⁷⁸ and its Connection Magic Local mode works “when there is no internet available,” through a portable router.⁷⁹ Stenograph’s CaseViewNet Cloud sends realtime over “any internet connection,” to people in the room or remote, and “Works with any CAT Software.”¹⁸ A practitioner’s checklist in NCRA’s journal lists a “Writer cable (when Bluetooth fails),” “2 microphones,” a “Power strip w/long cord,” and, for realtime jobs, a “Router” and a “Realtime/rough draft disclaimer.”⁸⁰

In person: voice

A voice writer in person runs two inputs and a backup. Stenograph’s VP guide says voice audio and room audio need two sound devices.¹² A two-output mask can feed the laptop and a separate recorder at the same time.¹³⁶⁴ Our reading: that recorder is your backup if the laptop fails. Check the seal and your level before the witness is sworn.⁶⁵ Our synthesis: borrow NVRA’s two-foot exam limit as the job’s quiet check, and ask someone at the table whether they can hear you dictate.⁷³

In person: digital

The digital setup is a microphone per speaking party, each on its own channel, monitored from the recording, with log notes as you go.¹⁵³ On a federal deposition recorded non-stenographically, “the officer must repeat the items in Rule 30(b)(5)(A)(i)-(iii) at the beginning of each unit of the recording medium,” and appearance or demeanor “must not be distorted through recording techniques.”²⁸ Our synthesis: treat each new file or card as a new unit. AAERT’s CDR exam counts administering oaths and managing exhibits as part of the job.⁵³ If your jurisdiction does not let a digital reporter take the deposition, check the rules first.

A placement recipe for a small conference table, DepoStack’s synthesis of the makers’ manuals cited, not an industry standard:

In person: on the record

NCRA Advisory Opinion 6 says “Unless all counsel in an examination agree to go off the record, the reporter is duty-bound to continue reporting,” must go on the record “immediately” if counsel disagree, and should tell the parties so before starting; it applies to members to the extent consistent with local law.⁸⁴ AAERT’s deposition exam tests “how and when to go on and off the record.”⁵³ The notice matters here too: Illinois requires an audio-visual notice to “contain the name of the recording-device operator,”⁸⁵ and Texas puts the party who wants a non-stenographic recording in charge of finding an oath-giver and making the recording “intelligible, accurate, and trustworthy.”⁸⁶

Services stay even. NCRA’s guidelines say “Court reporters should offer comparable services to all parties in a litigation proceeding,”⁴⁷ and California’s Board says a backup-audio copy offered to one party must be offered “at the same time” to the others.⁴⁶

In-person pack list. DepoStack’s synthesis of the sources cited, not an industry standard.

• The notice, and the recording method it names.²⁸
• Steno: the writer, charged, and its cable “when Bluetooth fails.”⁸⁰
• Voice: the mask with both outputs connected, and a second sound device.¹²¹³
• Digital: a mic per speaking party, plus a table mic or two.¹⁵
• A power strip with a long cord; for realtime, a router and a draft disclaimer.⁸⁰
• A battery backup recorder outside the main audio chain.¹⁶
• The off-the-record statement, said before you start.⁸⁴
• A privacy filter for the screen.⁸⁷

Remote depositions

Remote: authority and oath

Under the federal rules, the parties may stipulate, or the court may order, “that a deposition be taken by telephone or other remote means,” and “the deposition takes place where the deponent answers the questions.”²⁸ California’s Board advises finding out “where the deponent is located because the laws of that jurisdiction take precedence.”⁸⁸

The table is DepoStack’s synthesis of the rules cited, not an industry standard, current as of October 13, 2026.

Remote platform settings

Whoever controls the platform controls its recording. NCRA recommends “that the reporter be designated the host whenever possible,” to avoid unauthorized use of the recording feature.⁴⁷ An MDL deposition order required “a ‘virtual waiting room’ that allows the court reporter to admit only individuals authorized to attend” and a live attendee list.⁸⁹ Zoom’s security controls include locking the meeting, a waiting room, and suspending participants’ video, audio, and screen sharing.⁹⁶

Identify yourself and the room. California’s Board asks reporters to put their CSR number in their display name, stay on camera if possible, prefer gallery view, and request no virtual backgrounds; the appearance page “should indicate whether each attendee was remote or in person,” and a “technical difficulty” parenthetical is appropriate.⁸⁸ A magistrate judge’s sample protocol requires ID “legible on the video record” and says “Conversations in the breakout rooms shall not be recorded.”⁹⁰

Recording and bots. California’s Board says reporters “may want to remind participants” before going on the record that recording is not permitted unless noticed or agreed by all parties (CCP 2025.330(c)), and should watch for surreptitious recording.⁸⁸ Illinois says “No recording of a remote electronic means deposition shall be made other than the recording disclosed in the notice of deposition,”⁸⁵ and Massachusetts says “No person other than the operator/videographer and court reporter may record the deposition by video or audio means.”⁴⁵ NCRA’s July 2026 guidance warns that “Remote proceedings may implicate multi-state recording consent statutes,” and that AI notetakers may record “even if the platform’s microphone is muted”; its suggested script asks anyone using “AI, recording, or automated summarization tools” to disclose it.⁹⁷ Two NCRA ethics committee members wrote, not as a formal opinion, that if a party objects, “the reporter cannot proceed with the recording feature enabled.”⁹⁸ NCRA Advisory Opinion 44 says a reporter “may not act as both the videographer and the verbatim reporter for the same proceeding.”⁹⁹

When the connection drops. NCRA says to tell the parties on reconnection and “invite the parties to repeat what had been said,” and that chat is not in the record “unless one of the parties reads it into the record.”⁴⁷

What the platform records.

• Zoom local recording can save “separate audio files for up to 80 participants,” but phone participants “will be merged into one audio file.”²⁰ The host must record or grant permission, and an interrupted conversion after the meeting can leave files “corrupted and non-recoverable.”¹⁰⁰
• Zoom’s own pages disagree on cloud recording: the settings page offers per-participant files for “up to 200 speakers,” while the file-format page says per-participant tracks are “(computer recording only).” Test on your account.¹⁰¹¹⁰²
• A Zoom cloud recording “will only record the main meeting and none of the breakouts will be captured.”¹⁰³
• In Teams, “Guests can’t record meetings”; non-channel recordings go to the organizer’s OneDrive, and “By default, only meeting organizers and co-organizers can download or delete the recording.”¹⁰⁴
• Zoom’s noise suppression defaults to “Auto”; “High” removes “background speech from others”; high-fidelity music mode “disables echo cancellation and post-processing.”¹⁰⁵ No source tested how these settings affect a transcript, so we recommend none.

Remote capture, by method

Steno. Stenograph’s Direct Audio, in CATalyst 22 and later, records platform audio when you “select Stenograph Virtual Audio 1 Input as the output (speakers) device” in Zoom or Teams.¹⁷ Its files “will only play for users on CATalyst version 22 or higher,” including RealTeam; otherwise use Single Channel audio. For in-person jobs Stenograph warns against sending the live feed to a speaker, because “you would likely experience an echo or other feedback.”¹⁷ Direct Audio takes up to eight channels from a hardware mixer with MME or WASAPI drivers, not ASIO, in a proprietary .sgmca format.²³ Advantage’s 2021 guide uses Stereo Mix or Voicemeeter, warns “Stereo Mix will not record your voice,” and strongly recommends headphones with Voicemeeter.¹⁹ Its Eclipse 12 training describes built-in loopback recording “without third-party software”; we found no manual page for it.¹⁰⁶ Windows’ WASAPI loopback works only on shared-mode streams.¹⁰⁷

Headphones. Veritext, a national agency, warns its reporters that headphones plugged into the writing laptop can mean “you will lose your CAT software audio,” and suggests testing a USB splitter first.¹⁰⁸

Voice. We found no NVRA guidance for remote jobs. Our synthesis, not a published standard: capture platform audio into your voice CAT the same way; keep the mask as your dictation input and never select it as the platform’s microphone; speak to the participants through a separate microphone you can mute; and listen on headphones.

Digital. When you host or may record, Zoom local recording gives “one file for each participant”; ask participants to join by computer audio, since phone audio merges.¹⁰⁰²⁰ BlueLedge advises “not relying on the same audio signal or internet connection” for the backup recorder, for example a phone on cellular, and does “not recommend the use of Bluetooth speakers for digital reporters.”²¹

Remote realtime and exhibits

CaseViewNet Cloud and Bridge Mobile both serve viewers in the room and remote.¹⁸⁷⁸ Send the link only to parties who ordered it: NCRA says uncertified drafts “may only be distributed to ordering parties,”⁴⁷ and federal realtime likewise.³ For third-party captions in Zoom, the host enables the caption API token and copies it; “Every meeting and breakout room session has a special URL.”¹⁰⁹ Gear and dictionary prep are in Realtime Court Reporting.

Illinois excuses sending exhibits ahead if participants “are able to view the exhibits in real time.”⁸⁵ Veritext says “some remote platforms allow online access to documents for only 24 to 48 hours,” so download marked exhibits during breaks.¹⁰⁸ See The State of Exhibit Marking in 2026, and the exhibit number planner for numbering.

Remote connectivity

The Florida Bar’s guide says “A wired internet connection is preferable over a wireless connection.”¹¹⁰ Microsoft says Wi-Fi and VPNs are not necessarily designed for real-time media, and that Teams “prioritizes audio quality over video quality” when bandwidth runs short.¹¹¹

Bandwidth figures disagree because they answer different questions. Zoom recommends about 1.0 Mbps up and 600 kbps down for high-quality group video and 3.8/3.0 Mbps for 1080p; these are recommendations, not stated minimums.¹¹² One reporting agency asks its reporters for “a minimum bandwidth of 12.0 Mbps upload and 12.0 Mbps download,” about 12 times Zoom’s group-video figure.¹¹³ One is a platform’s recommendation; the other, an agency’s headroom.

An NCRA-published column says to buy “a Battery Backup & Surge Protector for your router and your laptop.”²² A Louisiana Judicial College paper, not a court rule, advises keeping “a smartphone hot spot or a telephone dial-in available in case your primary internet fails.”¹¹⁴

Remote pre-flight. DepoStack’s synthesis of the sources cited, not an industry standard.

• Where the witness is, and whose oath law applies.⁸⁸
• Host where possible, waiting room on, meeting locked once all are in.⁴⁷⁹⁶
• Display name with your license number; a roll for the appearance page.⁸⁸
• Recording reminder and AI-tool disclosure before going on the record.⁸⁸⁹⁷
• Platform audio routed into your CAT or recorder,¹⁷ and a test recording played back (our synthesis).
• Headphones tested with the CAT audio.¹⁰⁸
• Wired connection; router and laptop on a battery backup; hotspot or dial-in ready.^11022114
• Digital: backup recorder on a separate signal and connection.²¹

Hybrid depositions

In a hybrid room, more than one open device feeds back. An NCRA-published column describes one speakerphone carrying the room and every other device set to “leave computer audio”;²² Veritext’s reporter panel says “‘Leave Audio’ should be used anytime there is more than one device joined to audio conference and feedback is occurring.”²⁴ The rules pull the other way on log-ins: Illinois’ comment says everyone in the room “should separately participate in the videoconference” where possible, and Massachusetts requires that everyone with the deponent “must separately log in.”⁸⁵⁴⁵ Our reading: they fit together if everyone logs in for video and identification, and one device carries the room’s audio.

Stenograph documents a two-computer pattern: the reporter’s computer runs CATalyst and the meeting to record audio, and “the additional computer will run just the web conferencing app for the in-person participants.”²³ For digital reporters, Veritext’s panel says to “mic the in person participants and then also place a mic near the phone.”²⁴ Michigan’s court procedures say video-conference audio “should be fed directly into the recording system,” or a mic moved near its speaker.¹¹⁵

No source publishes a tested hybrid wiring diagram. This channel plan is DepoStack’s synthesis of the sources cited, not an industry standard, and untested: try it before a live job.

Videotaped depositions

Two records run side by side, and the roles stay separate: NCRA says a reporter “may not act as both the videographer and the verbatim reporter,”⁹⁹ and it “strongly recommends the use of a Certified Legal Video Specialist (CLVS).”⁸

On camera. Federally, a non-stenographic recording repeats the opening at each unit of the medium.²⁸ In Florida, “Audiovisually recorded depositions must also be recorded stenographically, unless all parties agree otherwise,” and the officer must, on camera, identify the style of the action, state the date, and swear the witness.²⁷ NCRA’s video standards log “on and off the record” separately from “on and off the video,” and have the videographer keep recording a disputed off-the-record request “until agreement by all counsel.”²⁶

Audio from the videographer. NCRA’s Standard #8 says the videographer “shall be able to provide the court reporter an audio recording which shall be the output from the audio mixer,” but not whether that is a live feed or a file, which connector, or what level.²⁶ Agree on it before the job. The standards also call for at least four microphones, three of them lavaliers, automatic gain off, and continuous headphone monitoring.²⁶

Timestamps. Stenograph’s “Set Time” format symbol is for when “you need to synchronize timestamps with a videographer.”²⁵ Advantage says Eclipse can match a separate digital recorder’s timecodes to the writer’s; that page addresses audio, not video.¹¹⁶ Sync formats and designations are in The State of Deposition Video Sync in 2026.

What you did not hear. NCRA Advisory Opinion 47: “A reporter should not include conversations unless they were actually heard and reported by the reporter,” except by written agreement of all parties, marked as a video excerpt.¹¹⁷

Voice writers near video mics. We found no source. Our synthesis: before going on the record, listen to the video track while you dictate at working level. If you are audible, fix the seal or move.

Court hearings and trials

Realtime to the bench

The federal program says court sessions “shall be recorded verbatim by shorthand, stenotype, stenomask, or electronic sound recording equipment,” as the district judge elects.¹¹⁸ Only certified realtime reporters (NCRA’s CRR or an equivalent exam) may sell realtime at Conference rates, and realtime is not required in all proceedings.³ Unless the court has a permanent router, the realtime reporter supplies the wiring, connectors, adapters, and router; parties bring “their own computers, viewer/annotation software, and monitors” and “should coordinate and pre-test their equipment with the realtime reporter”; and “Non-judiciary participants may not receive access to the DCN at any time.”³ Federal realtime is labelled “Realtime Unedited Transcript Only,” carries no appearance, index, or certification page, goes only to ordering parties, and may not be used where a certified transcript is required.³

Local courts write their own specs. Fresno County Superior Court’s mandatory standard does not support wireless realtime, requires a wired serial link with settings matched at both ends, and says court technicians do not directly support reporter-owned equipment; the details are in Realtime Court Reporting.²⁹ For a bench conference, know which writer link survives unplugging (Core kit: steno).

Federal official qualifications are set in a volume of the Guide that is not public. One 2026 posting in the Western District of Texas required NCRA’s RPR “or … an equivalent qualifying examination* such as the Certified Verbatim Reporter (CVR),” preferred realtime certification, and required the reporter to own realtime equipment; it survives only in an archive capture.¹¹⁹

Officials’ backup audio

The rules in Where audio becomes the record govern: personal property in federal court unless filed as the record of an arraignment, plea, or sentencing; never backed up by the court’s system; and in Minnesota, a court-owned backup is required.³⁵¹ California requires official reporters who keep notes in non-paper form to store “one duplicate backup copy” and review the media for obsolescence.¹²⁰

Digital in the courtroom

In court, a digital reporter usually runs the court’s installed system, and federal policy says it produces “No simultaneous text or transcript … beyond the recorder’s log notes.”³ The standards:

• Federal: four channels minimum (suspendable where not advantageous), channel isolation, monitoring from the recorded signal, log notes stamped when begun, non-proprietary export, an off-line backup plan, and emergency fallbacks.³
• California: at least four channels and four microphones that, absent unusual circumstances, “must be directional (cardioid),” with earphones on the recorded signal; the rule’s text still describes older media.¹²¹ The statute limits electronic recording to limited civil, misdemeanor, and infraction cases without a reporter and bars “an unofficial record”;¹²² since August 10, 2026, courts must give indigent civil litigants, on request, “meaningful access in all cases to an official verbatim record,” “including by means of electronic recording” if no reporter is available.¹²³ Contra Costa and Santa Clara then rescinded their electronic-recording general orders.¹²⁴¹²⁵
• NCSC: a default channel map (“1 - Judge/Jury/Bench or Well 2 - Witness 3 - Defendant 4 - Plaintiff”); the monitor should “Interrupt the proceeding and notify the Judge when a record is not being made,” using “best judgment”; and sidebars are recorded “unless the Judge orders otherwise.”³⁴
• Arizona: “Toggling mute buttons must not be used on microphones,” though mics that mute only while held are allowed, and systems must be capable of confidence monitoring.³¹
• Michigan: store the signal “to two separate storage devices simultaneously,” with “at least visual indication” that each microphone is picking up a signal.¹²⁶
• Florida: in the Eighteenth Circuit, digital court reporters monitor and “immediately signal or notify the trial judge” when quality is in question, and digital reporting is barred as the system in capital cases, as the statewide rule also requires.²¹²⁷
• Georgia: Rule 36.19 requires multichannel recording, digital backups, and “microphones on counsel tables that are individually mutable by the user”; it “does not apply to depositions.”⁵²
• Texas: an October 2026 study, not a rule, recommends “multi-channel audio capture with voice isolation, real-time confidence monitoring by a trained court recorder, time-stamped log notes, redundant backup of audio files, and a defined retention period.”⁵⁶

Daily and expedited copy

Federal daily copy is due “on the calendar day following receipt of the order,” before the clerk’s office opens, and hourly copy “within two hours”; those rates cover “extra court reporters, typists, and transcribers.”³ When the court cannot provide expedited delivery, a judge may allow a private reporter to make “an unofficial expedited record.”³ Stenograph’s RealTeam shares text and audio with scopists and proofreaders while you write; it needs CATalyst V20 or later, a support plan, and internet.¹²⁸ For due dates, use the turnaround date calculator.

Arbitrations, EUOs, IMEs, and meetings

The evidence for these jobs is thin. California’s definition of shorthand reporting reaches “any oral court proceeding, deposition, court ordered hearing or arbitration.”¹²⁹ AAERT’s CER description covers “depositions, EUOs, statements on the record, aids of execution, meetings, arbitrations,” while its CDR description lists EUOs and statements but not meetings or arbitrations.⁶ We found no setup guidance specific to these jobs. Our synthesis: your deposition setup carries over; settle at booking who hired you, what the record is for, and who gets copies.

Medical content and HIPAA

In California, an examinee’s attorney may “record stenographically or by audio technology any words spoken to or by the examinee,” and examiner and examinee may audio-record a mental examination; other states were not researched.¹³⁰¹³¹ A HIPAA qualified protective order must bar use outside the litigation and require “the return to the covered entity or destruction of the protected health information (including all copies made).”³² HHS says a lawyer who is a business associate must ensure “litigation support personnel, or others hired by the lawyer to assist the lawyer” protect health information; it does not name court reporters.¹³² Whether a given reporter is a business associate is fact-specific, so ask hiring counsel how the order applies to your notes, audio, and copies. Our reading: the copies to return or destroy include backups. California’s Board says to delete scanned documents “from all backup sources,”¹³³ and CATalyst’s cloud backup keeps files deleted in CATalyst.¹³⁴

Multi-day litigation

RealTeam shares text and audio with your team while you write,¹²⁸ but Direct Audio files play only on CATalyst 22 or later, so confirm your scopist’s version.¹⁷ NCSC suggests taking appearances “at the start of each day” of a multi-day trial and keeping “a separate word list with the spelling of proper nouns and technical jargon.”³⁴

Interpreted proceedings

California’s Board says “THE COURT REPORTER BEGINS BY SWEARING IN THE INTERPRETER” and the reporter “must make the record clear as to which language is used.”³³ A bilingual reporter “is not to interrupt to correct the interpretation,” but must interrupt if no record is being created, and “it is often very helpful to the interpreter to have access to a realtime screen.”³³ For voice writers, Talk Technologies’ manual describes “lagging,” waiting several words before repeating to cut false starts;⁶⁶ our suggestion is to use it with interpreted answers too. Transcript conventions are in The Art of the Transcript.

Modes. In federal proceedings instituted by the United States, interpretation is simultaneous for parties and “consecutive mode for witnesses” unless the judge orders otherwise, and on a party’s motion the judge decides whether to require an electronic sound recording.¹³⁵ Washington’s comment says consecutive mode for a witness “allows for the English interpretation to be on the record,” and requires consecutive interpretation when remote technology cannot carry simultaneous.¹³⁶ Iowa requires courts to record testimony given through an interpreter in a language other than English.¹³⁷

Channels. NCSC says the interpreter’s microphone goes on “a channel that is not the same as the channel assigned to the witness,” muted carefully “to ensure that a witness’s private conversations are not recorded”;³⁴ Michigan says the same.¹¹⁵

Remote. Zoom cloud recordings keep the original audio and, with a setting on, “an audio file for each language interpretation”; local recordings capture only what the recorder hears; and phone users cannot hear interpretation.¹³⁸ In Teams, “The interpreter’s audio will not be present on the recording,”¹³⁹ and Webex recordings “only capture the original audio.”¹⁴⁰ California’s remote-interpreting guidelines favor simultaneous-capable solutions and headsets, and do not recommend speakerphones; they do not address court reporters.¹⁴¹

CART and captioning

CART is a separate job from the official record. Federal policy says the CART provider “must not be the court reporter taking the official record,”³ and NCRA’s captioner code says to “Refrain from working in the dual capacity of official reporter of proceedings and CART captioner at the same time,” and to disclose it on the record if it cannot be avoided.¹⁴²

Files. In a legal setting, “No file is to be provided unless by order of the Court.”¹⁴² NCRA’s CART manual makes preserving the unedited file, with a disclaimer, for “no fewer than five years” the default, but in a confidential setting such as jury deliberations or attorney-client talk, the display “should not be visible to others” and the captioner “should delete all files immediately after the assignment unless requested not to do so or ordered by the Court.”³⁵

Kit and display. The manual’s kit includes an “additional realtime cable,” CAT with “text-enlarging capability,” a “surge protector with indicator light,” and gaffer tape, with “Backup equipment, computer, and software key/code” readily available. It says to test compatibility before agreeing to display on external equipment, to blank the display while making job defines, and that “Brackets should not appear around untranslated words.” For audio, it suggests pulling audio “off the audio mixer board with headphones”; for remote CART, a designated URL with “emergency contact capability.”³⁵

Backup and timing. NCRA’s 2016 guidelines say a captioner “should have a backup computer and steno machine,” name a UPS or generator as backup power, and say to be set up 30 to 60 minutes early on site, or 10 to 15 minutes early remotely.¹⁴³

The legal frame. Under ADA Title II, auxiliary aids include “real-time computer-aided transcription services” and “real-time captioning,”³⁷ and public entities “shall give primary consideration to the requests of individuals with disabilities.”¹⁴⁴ Federal court policy is in tension with that. Volume 5 of the Guide limits court CART to “a video display of spoken words,” “not to be used instead of conventional means of producing the official record,” yet says courts “should honor a participant’s choice of auxiliary aid or service” unless an equally effective alternative exists or the choice would be an undue burden.¹⁴⁵ Volume 6 says CART should be approved “only after alternatives that are more cost effective have been considered.”³ Volume 5’s exception partly reconciles the two.

By method. NCRA’s CRC is a five-minute literary test at 180 wpm and 96 percent plus a written test, for “current or aspiring stenographic captioners.”¹⁴⁶ NVRA describes CART as produced with “speech-to-text software, a computer, and either an open or closed microphone or stenotype machine”;¹⁴⁷ its RCP-M exam runs up to 225 wpm at 97.5 percent with a “standard drop down rate of 5 seconds or less,” taken on a “stable Internet connection, preferably hardwired.”³⁶ For digital reporters, no source we read treats a recording as CART. That is a gap in the sources, not a judgment; the ADA list names real-time transcription and captioning,³⁷ and federal court CART is a live display.¹⁴⁵ See also CART vs. deposition realtime.

CART booking sheet. DepoStack’s synthesis of the sources cited, not an industry standard.

• Who hired you, and that you are not also the official reporter.³¹⁴²
• File disposition, in writing: none without a court order in a legal setting; kept or deleted by setting.¹⁴²³⁵
• External display tested before you accept; display hidden during confidential talk.³⁵
• Audio from the room’s mixer or an assistive listening device.³⁵
• Backup computer, writer, software key, and power.¹⁴³
• Arrival 30 to 60 minutes early on site, 10 to 15 remotely.¹⁴³
• A signal, agreed with the consumer, for when text stops (our synthesis; no source sets one).

Realtime and rough drafts

The federal program defines realtime as “A draft unedited transcript produced by a certified realtime reporter as a byproduct of realtime.”¹¹⁸ Who can produce it is contested. NCRA’s position: “Stenographic reporters and voice writers can provide this service; digital recorders/transcriptionists cannot.”¹⁴⁸ Stenograph says its MAXScribe “enables digital reporters to offer realtime reporting,”¹⁴⁹ and For The Record says its RealTime Go “produces accurate, real-time transcripts.”¹⁵⁰ We did not verify those claims, and no source we read shows such text accepted as certified realtime. Federal policy: a digital audio system produces no simultaneous text beyond the log notes.³ If a client orders realtime from a digital reporter, tell the client before the job that the recording will not produce live text (our suggestion).

By method. NCRA’s CRR is a five-minute realtime Q&A at 96 percent, and candidates “cannot edit your transcript at all before uploading.”³⁸ NVRA’s RVR is one of three eight-minute segments at 180 wpm and 95 percent, the RVR-M an eight-minute 200 wpm Q&A at 96 percent, both “utilizing a closed microphone system.”⁴⁰⁵ Stenograph says CATalyst VP’s “Dragon-Faster-Drop” “puts the text on the screen faster than Dragon operating alone.”¹⁵¹ Gear and dictionary prep are in Realtime Court Reporting.

Rough drafts. NCRA says “An uncertified transcript may not be filed with the court,” “may only be distributed to ordering parties,” “may be provided in condensed format only,” and the signed disclaimer should be kept.⁴⁷ California’s Board allows a changed line count with labels instead, and says a draft should state it is “unedited and uncertified,” “should never include a completed title page, appearance, index, or certification page,” and should have “All untranslated steno strokes/undefined voice commands” resolved before delivery.³⁹ The two differ on format; both keep drafts apart from the certified transcript. California’s statute says a rough draft “may not be certified.”¹⁵² Offer the same services to every party.⁴⁷ For draft versus certified text in review, see The State of Scoping and Transcript Review in 2026. The transcript invoice estimator has a realtime per-page line.

Field and restricted sites

No power and travel

The FAA says spare lithium batteries and power banks “must be carried in carry-on baggage only,” limits rechargeable batteries to “100 watt hours (Wh) per battery” without airline approval (up to two spares of 101 to 160 Wh with approval), and forbids batteries over 160 Wh; airlines “may have stricter” limits.¹⁵³¹⁵⁴

Stenograph says its NexGen runs “approximately 12 hours and then will go into FailSafe mode which may last up to 3” hours.⁹ In FailSafe, the Luminex II keeps notes, audio, and USB realtime going with the screen and radios off.⁴¹ Our reading: on a long day without an outlet, USB keeps realtime alive longest. A writer powered over USB draws on the laptop’s battery instead.⁵⁹ Recorders lose runtime fast with load (see the battery table).⁴² Practitioner advice in NCRA’s journal: “Have two of everything,” and “Never count on an outlet being close.”⁸⁷

Jails and prisons

The Bureau of Prisons limits a legal visitor’s power cord to six feet, bars recording “Unless specifically authorized in advance,” requires wireless off or airplane mode, and prohibits connecting to its network.¹⁵⁵ The Los Angeles County Sheriff’s laptop form says the department “is not obligated to provide a power source” and bars phones and hotspots; it names attorneys and certain representatives but not court reporters.¹⁵⁶ California’s prison system requires the attorney’s request to include the reporter’s date of birth, Social Security number, and driver’s license number, and makes the attorney responsible for telling the reporter the entry requirements.¹⁵⁷ A federal deposition of a person “confined in prison” requires leave of court.²⁸ We found no published hospital or bedside equipment policy.

No-outlet day. DepoStack’s synthesis of the sources cited, not an industry standard.

• The writer fully charged, its FailSafe behavior known.⁴¹
• A spare SD card.⁴¹
• Laptop power settings checked before you leave.¹⁵⁸
• Recorder runtime budgeted for your actual load.⁴²⁴³
• Spare batteries in your carry-on, 100 Wh or less each unless approved.¹⁵³
• Two of everything essential, chargers first.⁸⁷
• For a prison: entry requirements from the attorney, approval for each device and any recording, and a short cord.¹⁵⁷¹⁵⁵

Audio capture

Microphones and placement

NCSC warns that omnidirectional mics “turn a sophisticated multi-channel recording system into a mono recording system” and that cardioid is the most common directional choice.³⁴ California’s court rule requires microphones that, absent unusual circumstances, “must be directional (cardioid).”¹²¹ Yet “Most lav mics feature an omnidirectional capsule,”⁸³ and NCRA’s video standards still call for three lavaliers.²⁶ Placement figures from the makers are in the in-person placement table. NCSC adds: “Request that a person stop tapping a microphone or shuffling papers too close to it.”³⁴

Channels

AAERT’s white paper explains why isolation matters: “By isolating the recording to one channel at a time, the transcriptionist can listen to one speaker at a time to determine what was spoken and by whom.”¹⁵⁹ Federal policy sets four channels as the courtroom minimum, suspendable where not advantageous, such as telephone proceedings.³ AAERT’s blog says a jury trial needs at least four if channels can be shared, and “Miking each location individually would require 8-9 channels!”¹⁵ Isolation can be lost downstream: Zoom merges phone participants into one file,²⁰ Amazon’s transcription service says “Media with more than two channels is not currently supported,”¹⁶⁰ and Google advises sending each speaker’s channel “separately.”¹⁶¹

Levels and processing

Zoom Corporation’s H6 manual says “Adjust so that the peak level stays around −12dB”; TASCAM’s DR-40X manual says to keep the meter “centered on -12 without causing the PEAK indicator to light red.” One is a peak, the other a meter centre.¹⁶²¹⁶³ AAERT puts good levels “between -12 dB and -6 dB” and says above 0 dB “your audio will clip.”¹⁵ Both recorders offer a quieter safety file: the H6 only for its L/R input, 12 dB lower; the DR-40X’s DUAL REC at −6 to −12 dB. On the H6, “When set to MP3, a stereo mix will be recorded regardless of the number of tracks.”¹⁶²¹⁶³ Zoom Corporation’s H6essential manual says its 32-bit float design records “both loud and quiet sounds without making gain adjustments”; Windows needs a driver to use it as an interface.¹⁶⁴ Our reading: that protects the converter, not a badly placed mic.

The sources disagree on automatic gain and noise processing, because their goals differ. NCSC says court systems must provide “Noise filtration and automatic gain control”; NCRA’s video standards set automatic gain “to the ‘off’ position”; and Google says “Don’t use automatic gain control (AGC)” and “All noise reduction processing should be disabled” for recognition.^3426161 For voice writers, Stenograph says to turn Dragon’s automatic level off.¹²

Formats

Federal policy has “no approved file formats,” names WAV as “non-proprietary,” and requires non-proprietary export.³ Arizona lists “AVI, MPEG/MP3, FLAC, or WAV playable by non-proprietary readers.”³¹ Florida’s court standard sets “a minimum sampling rate of 44.1KHz at 16 bits,” and is internally inconsistent on whether a courtroom needs eight channels or four.¹⁶⁵ Google asks for 16,000 Hz or higher and “a lossless codec.”¹⁶¹ CAT defaults can sit lower: Case CATalyst’s default codec is “ADPCM 8000 Hz,” with “PCM 48000 Hz” available at much larger sizes.¹⁶⁶ Our reading: if anyone else will work from your audio, check its format and rate before the job.

Monitoring

Listen to the recording, not the room. Federal monitoring uses “the already recorded signal … (not the signal directly from the microphones)”;³ Michigan’s manual says the operator “listens to the recording, not the source”;¹¹⁵ and Michigan’s standards require “at least visual indication” that each mic is live.¹²⁶ AAERT says without “confidence monitoring,” “it’s impossible to detect and correct any technical issues.”¹⁶⁷ BlueLedge says never to use noise-cancelling headphones in a proceeding.¹⁶

Recording laws apply to everyone’s audio. California’s Board cites “California Penal Code 632, which prohibits eavesdropping on or recording confidential communications.”⁴⁶

Power and redundancy

Writer and recorder batteries

Makers’ own figures under the conditions they state, not our tests.

Laptop and CAT settings

Stenograph tells CATalyst users to set Windows to Best Performance on battery and plugged in, set sleep and lid actions to Never, and stop Windows from powering down USB root hubs.¹⁵⁸ Autosave differs by program. Case CATalyst saves at a user-set interval, which Stenograph recommends be at least five minutes, and keeps the original after an abnormal end (guide version 18, 2017).¹⁶⁹ Total Eclipse “saves to disk each time the cursor moves out of a paragraph,” with timed auto-backup.¹⁷⁰ ProCAT’s Autosave “only makes a copy” in a .SAV file.¹⁷¹ StenoCAT updates “your transcript and audio in both locations.”¹⁷² We found no documentation of DigitalCAT’s transcript autosave.

Backups and retention

Stenograph recommends storing backups “off of your computer like an external hard drive or the cloud.”¹⁷³ CATalyst Cloud Backup pauses “during realtime,”¹⁷⁴ and files deleted in CATalyst stay in the cloud backup.¹³⁴ For recordings, Michigan requires storing the signal to two devices at once,¹²⁶ and Arizona requires daily backups, saying “Simultaneous storage is not a substitute.”³¹

Retention follows the governing law. California requires deposition notes to be kept at least eight years when no transcript is produced, or one year after a transcript is produced.¹⁷⁵ The bodies’ defaults differ: NCRA “no less than five (5) years,”⁴⁷ AAERT “no less than three years” for recordings,¹⁷⁶ and NVRA defers to “rule, statute or court order.”¹⁷⁷ Advantage’s audio page assumes “you’ll delete the sound files when you’re finished transcribing,”¹¹⁶ which conflicts with NCRA’s guidance that backup audio be preserved on request “for the same period of time for which the reporter’s notes are preserved.”⁴⁷

Two of everything. DepoStack’s synthesis of the sources cited, not an industry standard.

• Notes in two places, plus a copy off the computer.⁴¹¹⁷³
• A main recording and an independent battery recorder.¹⁵¹⁶
• USB and Bluetooth paired for the writer.¹⁰
• Charged batteries, and a battery backup for router and laptop.²²
• Wired internet, and a hotspot or dial-in number.¹¹⁰¹¹⁴
• Spare chargers, cables, a power strip with a long cord, and tape.⁸⁷⁸⁰³⁵
• Autosave on, at an interval you have checked.¹⁶⁹

Security and confidentiality

Codes and rules

NCRA’s code requires members to “Preserve the confidentiality and ensure the security of information”;¹⁷⁸ AAERT’s uses nearly the same words and adds “Do not discuss remarks heard on or off the record with anyone”;¹⁷⁶ NVRA’s says “Preserve confidentiality; respect privileged information.”¹⁷⁷ NCRA’s position on outsourcing says accountability for security and confidentiality passes with the work “to the scopist, proofreader, transcriptionist, production facility, or whomever.”⁸ California makes confidentiality an enforceable standard for licensees.¹⁷⁹

Device and file security

California’s Board advises strong passwords on external drives, “multifactor authentication (MFA) whenever possible,” backups to “encrypted cloud services or password-protected drives,” and sending transcripts and exhibits by “encrypted email or file transfer services”; deleting scanned documents “from all backup sources” when done; a retention policy; and wiping devices before disposal. It is guidance, not a basis for discipline.¹³³ Microsoft says Windows device encryption “doesn’t encrypt external/USB drives,” and a device with only local accounts “remains unprotected even though the data is encrypted.”¹⁸⁰ Federal policy asks for current antivirus, passwords, and offsite backups, and its digital-audio security plans cover “sending unencrypted files over the internet” and “unauthorized listening to sealed or other sensitive court proceedings.”³ Florida’s court systems must support sealing with “user authorization, encryption, and seal keys,” and NCSC calls for “record-over protection.”¹⁶⁵³⁴ AAERT’s exams test “the chain of custody of the digital reporter’s audio, job information, and exhibits.”⁵³¹⁸¹ At the table, one practitioner’s answer to a lawyer reading over her shoulder was a privacy filter.⁸⁷

AI tools and bots

NCRA’s AI guidance says “There must be no unauthorized transcript uploading, data sharing, storage, or reuse of proceeding content,” and, on summaries, “Do not input prompts or analytical instructions into AI tools.”⁵⁴⁹⁷ Read a cloud service’s terms first: Amazon’s transcription service, for one, “may temporarily store your content to continuously improve the quality of its analysis models.”¹⁶⁰ For bots that join a remote proceeding, see Remote platform settings.

Releasing audio

If you release backup audio, NCRA says “release a copy and not the original (unless ordered otherwise by a court),” make the same offer to every party, and make sure no confidential or off-the-record talk is in it.⁴⁷ California’s Board says backup audio for a sealed or confidential transcript should be “clearly labeled,” and the original kept “for no less than one year.”⁴⁶

Rules and credentials

Rules by jurisdiction

As of October 13, 2026. Each cell cites the rule it states. “Not addressed in the sources” means the sources we read did not address it; it is not a finding that a method is barred. Rows marked in flux have a pending case, a new rule, or a recent bill.

NVRA says voice writers may practice in the state courts of most states and marks Montana, Vermont, and Wyoming “Freelance only,” but its two published lists do not match and its state-requirements PDF no longer loads; it also says “The rules for Federal Courts allow for machine writers, voice writers, and digital recorders.” Verify state by state.²⁰⁰⁵

Credentials by method

Positions of professional bodies

NCRA. Its president wrote in 2023 that “a stenographic court reporter is the best means of ensuring that the record is protected.”²⁰⁷ Its 2023 white paper says “there is no universal standard” for digital recording equipment and that transcripts by someone other than the recorder “make chain of custody difficult to track”;²⁰⁸ federal policy and Washington’s GR 35 expressly provide for separate transcribers.³¹⁹⁶ An NCRA advocacy flyer says recordings “can be riddled with inaudible statements” and that “The only requirement for digital recorders is a few hours of training and becoming a notary public.”¹⁴⁸ That is NCRA’s position. It conflicts with AAERT’s certification exams, whose CET practical is a four-channel mock proceeding scored at 95 percent,⁶ and with Tennessee’s licensing, which licenses electronic reporters on AAERT’s CER and CET.¹⁹³

NVRA. “This standard is equivalent to those set by the National Court Reporters Association (NCRA) for stenographers.”²⁰⁹

AAERT. It distinguishes certified digital reporting from unmonitored recording.¹⁶⁷ It calls professional oversight, not the tool, “The true gold standard.”⁵⁵ Its 2025 industry report urges “regulatory updates that remove barriers to digital reporting” and reports stenographers down 21 percent in a decade, to about 23,000; those are AAERT’s figures, and we did not review its methodology.²¹⁰

Others. The Speech-to-Text Institute’s archived pages say “no single technology holds exclusivity on such quality”; its current status is unverified.²¹¹ Texas’s 2026 survey found about 8.9 percent of court reporters, against about 69 percent of court recorders, agreed that digital recording is as accurate as a stenographic reporter; that measures belief, not accuracy.⁵⁶ And the California Supreme Court declined to compare: “a verbatim record created by electronic recording is better than no record at all.”¹²³

In flux

• Texas. The Supreme Court of Texas heard argument in In re Hughey on October 6, 2026, on whether a trial court abused its discretion in striking a non-stenographically recorded, AI-generated deposition transcript; as of October 13, 2026, it was “Submitted (Awaiting Opinion).”⁵⁷ Below, the trial court struck the transcript but not the video, and the court of appeals denied mandamus without deciding the statutory question.²¹² The 2023 law that changed “recorded” to “reported” includes an uncodified clause saying its Article 10 “is and shall be construed to be consistent with” the rules allowing non-stenographic recording.²¹³
• California. Los Angeles’s and San Diego’s responses to the August 10 decision were not found.¹²³
• Georgia. Rule 36.19 took effect February 26, 2026,⁵² and the scope of an AAERT-based Georgia certificate is not stated.¹⁸⁹
• Louisiana. HB 188 was deferred in committee on March 26, 2026.¹⁹⁹
• New York. S10374 was “In Senate Committee.”⁹⁴
• AAERT. A 2026 Best Practices Guide is announced,⁷⁷ and the CAPTUR merger was ratified in 2026.²⁰³
• Minnesota. A revision of the backup policy after the 2024 text likely exists.⁵¹

What we could not verify

• Dragon 16’s lifecycle: no end date on Nuance’s pages.⁷⁰
• Speechmatics data terms in Advantage’s published terms.⁷⁴
• Windows 11 support for Talk Technologies or Martel masks and the Andrea USB adapter.⁶⁷
• Zoom cloud per-participant audio: Zoom’s pages disagree.¹⁰²
• Florida and voice writing under Rule 1.310.²⁷
• Georgia’s AAERT-based certificate: its scope of practice.¹⁸⁹
• Hospitals: no published equipment policy; jails beyond three sites not researched.
• AAERT’s Best Practices Guide: the paid 2019 edition was not reviewed.⁷⁷
• Reporters under HIPAA: no HHS statement on whether a reporter is a business associate.¹³²
• Platform audio processing: no source tested its effect on transcripts.¹⁰⁵
• Hybrid wiring and remote voice writing: no tested diagram, and no NVRA guidance.
• NCRA Standard #8: feed, connector, and level unspecified.²⁶
• Federal officials: whether NVRA’s RVR counts as an equivalent realtime exam.¹¹⁹
• Tennessee: the effect of a 2025 law restructuring its board; the statute was read on FindLaw.¹⁹²
• One writer’s battery figure: its maker gives two numbers, so it is not in the table.

Build your checklist

The setup picker turns this guide into a job sheet: choose your method and job, and print a checklist where every line links back to the section that sources it.

The gear changes with the method and the job. The signature does not. The reporter or the transcriber certifies the record; the software never does.