Verbatim AI transcription built for
Australian courts and tribunals.
Nguma produces certified, human-reviewed transcripts for civil, criminal, and appellate proceedings, tribunal hearings, coronial inquests, Royal Commissions, and mediations. Australian-hosted. Chain-of-custody logged. Court reporting conventions applied per jurisdiction. No offshore processing. No model training on your recordings. At our volume tier, approximately 58% less than published conventional rates.
Transcription is critical infrastructure. It deserves better tools.
Whether the room is a courtroom, a tribunal, a commission, or a mediation, the record matters. The tools available to produce it have not kept pace with what courts, practitioners, and parties expect.
Turnaround that does not fit how work happens now
Conventional transcription was built for a different pace. Standard turnaround across Australian court transcription providers runs 3 to 7 business days. A 24-hour transcript is an express premium tier at every provider that offers it. Nguma delivers to authorised reviewers as soon as AI processing and human sign-off are complete — typically the same day, at a flat hourly rate with no express surcharge.
Sovereignty and confidentiality exposure
Mainstream AI transcription tools route audio through overseas servers and train models on customer data by default. Recordings of suppressed proceedings, board deliberations, and protected interviews cannot leave Australia. APP 8 of the Privacy Act 1988 is engaged on every upload to an offshore tool, whether the user is aware of it or not. For court content, this is not a configuration risk; it is a fundamental misalignment between the tool and the obligation.
The cost of the status quo
Per-word transcription pricing is the industry norm, and it scales with how much is said rather than with the value of the record. Across the roughly 150 Victorian courtrooms sitting every working day, the annual cost runs well into the millions, before tribunals, inquiries, parliamentary committees, and corporate proceedings are added.
The same court day. 58% less.
Nguma is priced in hourly bundles of audio rather than per word. Bundles are pre-purchased at the start of the engagement; the effective hourly rate decreases as bundle size increases.
A court day under the conventional model
Per-word billing scales with how much is said rather than with the value of the record, and a single week of proceedings across one courtroom adds up quickly. For jurisdictions running dozens or hundreds of courtrooms simultaneously, the annual figure is structurally large regardless of efficiency improvements elsewhere in the courts budget.
- Scales with how much is said, not with the value of the record
- Turnaround measured in days or weeks for some jurisdictions
- Budget difficult to forecast accurately across complex matters
Source: Court Services Victoria published transcript rates. Rates vary by jurisdiction; the CSV rate is used here as a documented publicly available benchmark.
The equivalent day on an hourly bundle
Nguma’s hourly bundle model delivers the equivalent court day at up to 58% less than conventional per-word rates. Each hour of audio includes scheduling, AI transcription, output to your nominated specification, human review, and the full chain-of-custody audit trail. The effective rate decreases as bundle size increases, making large-volume engagements proportionally cheaper.
- Predictable hourly rate, forecastable budget
- Draft transcript within hours; certified same day or next day
- Volume discounts apply across the engagement
Indicative rate based on published CSV benchmark. Exact pricing provided on brief once volume and output specification are confirmed. Contact us for a quote.
Get a quote for your jurisdiction.
Tell us the proceeding type and volume. We will come back to you within one business day.
What your legal and procurement teams will ask about.
Procurement of transcription services for courts and tribunals is a risk decision before it is a purchase. Nguma is built from the ground up to satisfy legal, security, and information governance teams.
Admissibility and evidentiary standard
Transcripts are produced from the certified audio record, time-stamped against that record, and reviewed by a human transcriber against the original audio before sign-off. Chain-of-custody is logged from upload to final transcript. Court reporting conventions are applied per jurisdiction. Final admissibility is always a matter for the presiding judicial officer; Nguma is designed to meet the evidentiary standard expected of certified transcripts in Australian courts.
Suppression order handling
Suppression and non-publication handling is configured at the matter level. Restricted content can be redacted from public-facing transcripts automatically, access limited to named users, and a controlled version produced for the requesting authority while the public record is suppressed or pseudonymised. Every access event is logged in the audit trail. In-camera content is handled under the same matter-level access controls.
Australian data residency
All audio, transcripts, and AI inference are processed and stored in Australia, with the application database and records held in AWS’s Sydney region and speech-to-text processed in Australia. Nguma does not transfer customer data offshore for storage in the ordinary course of the service. Full architecture and data-residency documentation is available under NDA for security assessment. Recordings of suppressed proceedings and in-camera content cannot be routed offshore regardless of how the system is configured.
Court audio never used to train AI
Nguma contractually guarantees that customer audio, transcripts, and metadata are never used to train AI models, fine-tune models, or improve the Nguma service. This is a hard contractual commitment reflected in standard terms. For court content, this is one of the central reasons Nguma exists as an Australian-sovereign alternative to mainstream AI transcription tools, where model training on customer data is typically opt-out rather than prohibited.
Human-in-the-loop at every step
AI produces the draft transcript; a human reviewer checks it against the original audio and approves it before it becomes the certified record. The AI never produces a final output without human review. For court transcription, this means every word in the certified transcript has been verified by a human against the audio, with confidence ratings flagging passages that warrant closer attention.
Information security and AI governance
ISO 27001 certification is in progress via Vanta, covering data classification, access management, incident response, and business continuity. ISO 42001 certification is also in progress, covering the AI governance framework including model selection, bias assessment, and human-in-the-loop controls. Both certifications are actively pursued alongside Essential Eight ML2.
Every court, tribunal, and commission in Australia.
Court reporting conventions, jurisdictional formatting, and suppression rules are configured per matter. Output is tailored to the specific requirements of the proceeding, not produced from a fixed template.
Federal and appellate courts
Federal Court, Full Federal Court, Federal Circuit and Family Court of Australia, High Court matters. Full court reporting conventions per jurisdiction.
Supreme, District, and Magistrates Courts
All state and territory Supreme Courts, District Courts, and Magistrates Courts. Civil, criminal, and appellate proceedings across all jurisdictions.
Tribunals
Administrative Review Tribunal (ART), NCAT, VCAT, SACAT, QCAT, SAT, and TASCAT. Jurisdictional formatting and member attribution applied per tribunal.
Coronial inquests
Trauma-sensitive content with controlled access. Suppression and restricted publication handling. Coronial conventions applied; witness and counsel attribution throughout.
Royal Commissions and inquiries
High-volume daily transcription with rolling suppression workflows. Counsel, witness, and Commissioner attribution. Real-time or same-day delivery to match publication timelines.
Mediations and arbitrations
Confidential dispute resolution with without-prejudice handling. Verbatim for formal arbitrations; structured summary for mediation sessions. Access restricted to named parties and representatives.
You set the spec. We deliver to it.
No fixed templates you must choose from. If you have an existing house style, court reporting template, or Hansard convention, we produce to that specification. If you would prefer us to recommend a starting point, we will.
Court-grade verbatim
The strictest output requirement: every word, hesitation, and non-verbal marker. Speaker attribution to court reporting standard.
- True verbatim including hesitations and false starts
- Speaker attribution to court reporting standard
- Time-stamped to audio cues throughout
- Suppression and non-publication handling configured
- Human-certified against original audio
- Chain-of-custody log included
Structured Hansard-style
Speaker-attributed, lightly edited for readability. Suitable for parliamentary committees, statutory inquiries, and integrity body interviews.
- Member and witness identification throughout
- Section headings and question-and-answer formatting
- Lightly edited for readability, not verbatim
- Suitable for Hansard and committee publication
- Human-reviewed and sign-off ready
- Exports to Word, PDF, or your publishing system
These are common starting points, not constraints. If your jurisdiction or house style requires something different, describe it when you brief us and we will produce a custom output specification for the matter. Submit your spec now
From recording to certified transcript in four steps.
Nguma works with the recording infrastructure you already use: in-room court recording systems, Microsoft Teams, Zoom, Webex, and direct audio file upload. No replacement of existing infrastructure required.
Configure the matter
Set the output specification, participants, suppression rules, access controls, and retention policy. Configure once per matter; recurring sittings use the same configuration automatically.
Capture the audio
Upload directly from your court recording system, connect to Teams or Zoom, or transfer via secure audio file. All audio is processed in Australia on receipt.
Transcribe and review
Australian-hosted AI produces a draft transcript within minutes. Your reviewer checks it against the audio using the built-in editor, which flags low-confidence passages and plays the audio in context. Edits are captured in the audit trail.
Sign off and export
Approved transcripts export to PDF, Word, JSON, or your case management system. The complete chain-of-custody log is preserved. Every action from upload to certification is timestamped and attributable.
What your procurement and legal teams will verify.
Questions your legal and procurement teams will ask.
Is Nguma transcription admissible as a court record?
Yes, when delivered with verbatim certification. Transcripts are produced from the certified audio record, time-stamped against that record, and reviewed by a human transcriber against the original audio before sign-off. Chain-of-custody is logged from upload to final transcript. Court reporting conventions are applied per jurisdiction. Final admissibility is always a matter for the presiding judicial officer, but Nguma is designed to meet the evidentiary standard expected of certified transcripts in Australian courts.
Where is the audio and transcript stored? Does any of it leave Australia?
All audio and transcripts are stored and processed in Australia, with the application database and records held in AWS’s Sydney region and speech-to-text processed in Australia. Recordings of court proceedings, statutory interviews, and investigations are never stored or processed outside Australia. Nguma does not transfer customer data offshore for storage in the ordinary course of the service. Full architecture and data-residency documentation is available under NDA for security assessment.
How are suppression orders and non-publication orders handled?
Suppression handling is configured at the matter level before the session begins. Restricted content can be redacted from public-facing transcripts, access limited to named users only, and a controlled version produced for the requesting authority while the public record is suppressed or pseudonymised. Every access event is logged in the audit trail, providing a full record of who accessed the restricted content and when.
Is customer audio or transcript content used to train AI models?
No. Nguma contractually guarantees that customer audio, transcripts, and metadata are never used to train AI models, fine-tune models, or improve the Nguma service. This is a hard contractual commitment reflected in standard terms, not a default setting that can be changed. For court content, this is one of the central reasons Nguma exists as an Australian-sovereign alternative to mainstream AI tools, where model training on customer data is typically opt-out rather than prohibited.
How does Nguma compare on cost to conventional transcription?
Nguma is priced in hourly bundles of audio rather than per word. At our volume tier, an equivalent court day costs up to 58% less than conventional per-word rates. Each hour of audio includes scheduling, transcription, output to your nominated specification, human review, and the full chain-of-custody audit trail. A quote is provided on brief once volume and output requirements are confirmed. Contact us to discuss your matter.
Can Nguma integrate with existing court recording systems?
Yes. Nguma works with existing court recording systems via secure audio file transfer, and natively with Microsoft Teams, Zoom, and Webex for hearings conducted via those platforms. Direct integration with case management and document management systems is available via API. There is no requirement to replace any existing recording infrastructure.
Can Nguma transcribe non-English proceedings?
Yes. Nguma can transcribe across a wide range of languages, with custom vocabulary support for legal and jurisdictional terminology. For First Nations and Australian Aboriginal languages, we engage qualified human translators in addition to AI processing, given the cultural and linguistic considerations involved. Please discuss specific language and cultural requirements when you brief us.
Get a transcription brief for your matter.
Tell us the proceeding type, volume, and output requirements. We will provide a custom output specification and a quote within one business day.