01
The litigator
Thirty hours of hearing and deposition audio across nine files. Know exactly what a witness said, when — and quote it with the timestamp behind it.
Recorded evidence intelligence
Built for legal matters, investigations, research, and high-stakes recorded archives.
Find what matters across hours of recorded evidence.
Upload audio, video, PDFs, Word files, spreadsheets, email exports and Outlook messages, ZIP archives, text and scanned images in bulk. Ask questions across the complete matter, see where sources disagree, what corroborates what and what rests on one source only, and trace every point to the exact recording timestamp or document page.
Thirty minutes free. No sales call required.
- Cited answersEvery answer links to the exact timestamp or page it came from.
- Human-in-the-loopFindings stay proposals until a person reviews and accepts them.
- English and SpanishTranscribe and question a matter in either language, with the interface to match.
- Audio, video and documentsOne matter, every format, searched together rather than app by app.
01
From recording to evidence
Why a transcript on its own is not evidence, and what changes when every line is citable.
A cited answer, not a guess
Ask the whole matter a question. Get an answer you can stand behind.
Every substantive claim resolves to an exact recording timestamp or document page — and when the evidence doesn't answer, the product says so instead of guessing.
When the material doesn't answer, the response is an honest “not enough here” — never an invented fact.
Illustrative example
Q
What did each participant say about the March meeting?
The claimant places it on the 3rd ¹; the respondent recalls “early March, maybe the 8th” ². Accounts differ on the date — flagged for review, not resolved.
Sources
- 1
call_09.mp3 · M. Ríos · t=47:12 - 2
interview_03.wav · J. Vale · t=12:04
The problem
The recordings hold the answer. Nobody can find it.
A single matter can carry dozens of hours of hearings, interviews and calls. The finding that matters is buried somewhere in minute 47 of file nine — and when someone finally quotes it, there's no clean line back to who said it, when, and in which recording. Transcription alone doesn't solve this: it just turns hours of audio into hours of unsearchable text.
The matter, today
- hearing_01.mp402:14:07
- deposition_02.m4a01:39:52
- interview_03.wav00:47:20
- call_09.mp3 · min 47:12◆“…on the 3rd, not the 8th.”
- + 5 more files06:11:44
The insight
People don't need a transcript. They need the record, made citable.
A transcript only converts audio into text. The useful step is connecting every recording in one searchable matter: ask across files, compare accounts, and open the exact speaker turn and timestamp behind each answer. That creates work you can verify and cite — not another document to read from beginning to end.
A transcript is
audio rendered as a long document. It can be read, but still leaves people searching file by file.
A record is
every recording connected in one searchable matter, where answers open back to the exact speaker turn and timestamp.
02
How the workspace works
What happens to an upload before you can use it: who spoke, what was said, and when.
Subject resolution
The hard part isn't the words. It's the person.
Speaker labels are local to each recording and diarization depends on the configured transcription provider. When someone explicitly introduces themselves, Tracecanon can propose that name for review; confirmed names link to one matter participant without pretending that an unverified voice match is certain.
01
Review a suggestion
Explicit self-introductions can produce a proposed name. A person confirms it before the label becomes an identity.
02
Link a participant
The same confirmed name maps speaker labels to a matter participant; no cross-customer voice matching is performed.
03
Keep uncertainty visible
Unconfirmed speakers remain neutral labels, and answers preserve uncertainty instead of silently assigning a person.
Identity is proposed for your confirmation, never assumed — the same human-in-the-loop rule as every finding.
The pipeline
From raw recording to cited work product.
One durable pipeline. Every stage writes to the database, not an in-memory queue — so a crashed worker's job is reclaimed and re-run, and you're charged only for processing that actually delivers.
- 01
Ingest
Direct-to-storage upload; validated by its bytes, hashed, probed.
- 02
Extract
Transcription for recordings; native text or OCR for documents and scans.
- 03
Index
Full-text + semantic vectors, tenant- and matter-scoped.
- 04
Cite
Ask the matter; every claim resolves to a recording timestamp or document page.
- 05
Analyze
Compare accounts, build chronology — proposed for human review.
- 06
Export
Professional formats + a ZIP bundle with a hash manifest.
03
What you can do
Ask questions across the whole matter, compare what people said, and export what you find.
What it is
The whole matter, made answerable.
Bulk upload that survives the browser
Drag in hundreds of recordings and documents, or whole folders. Uploads go straight to secure storage, resume after interruptions, and process in the background.
Transcription & document extraction
Recordings receive timestamps; PDFs, Word files, text, and scanned images retain page-level citations; spreadsheets cite the sheet and cell. Email attachments and ZIP contents become items of their own. Speaker separation is provider-dependent.
Search & cited answers
Ask a question across the whole matter. Every substantive answer cites an exact recording timestamp or document page — and says so when the evidence isn't there.
Cross-evidence comparison
Compare accounts across recordings and documents and surface potential inconsistencies — always shown with both sources and context, for human review.
Chronologies
Build a timeline of dated events pulled from the record, traceable back to source.
Professional exports
Polished PDF and DOCX transcripts, SRT/VTT captions, evidence bundles — with verification legends and, when enabled, hash manifests.
Evidence safeguards
Source hashes, upload manifests, verification status, and an audited trail. These record provenance and processing history — they don't claim authenticity or admissibility.
English & Spanish
Interface, transcription, cross-language search (ask in Spanish about English sources and vice versa), and exports — both languages from launch.
Security & privacy
Tenant isolation, private storage, signed URLs, a malware scan of every file before processing, two-step verification (required for our admins), and an append-only audit trail. Your matters are never mingled with other customers or products.
Who it's for
Anyone whose truth lives in hours of recordings.
02
The investigator
A dozen recorded interviews. Compare accounts across sessions and surface where two statements may not line up — carefully, defensibly.
03
The researcher
Years of interviews in two languages. Search the whole archive by meaning and cite every quote back to its source recording.
04
Review, provenance, and control
Every finding is yours to accept or reject, and every one shows the recording it came from.
Comparison, done responsibly
It surfaces potential inconsistencies — never verdicts.
When two statements from different recordings seem to differ, the workspace flags them for a human to review. It never says anyone lied — it accounts for corrections, hypotheticals, different time periods and translation, and keeps every finding proposed until a person confirms or dismisses it.
12
labels it may return — none can assert dishonesty
0
findings shown as settled — every one is human-reviewed
The only labels it may return
Potential inconsistencyNumerical discrepancyDate / timelineAttributionTextual differenceExplained by contextInsufficient evidenceNot inconsistentRequires human review
Amber flags something to look at · green clears it · grey is the tool declining to answer.
Anything unrecognized collapses to “requires human review.” Confidently-clean pairs are dropped so only what's worth a look survives.
Chronology
What happened, in what order — traceable to source.
Every dated statement across recordings, documents and email assembles into one timeline, with clock times, who says it, and one click back to the moment or page. Where sources give different times for the same happening, both versions stay visible and are linked to the finding that compares them.
Matter timeline
Mar 3
Meeting held, per the claimant
call_09.mp3 · t=47:12Mar 8
“Early March” recalled by respondent
interview_03.wav · t=12:04Apr 14
Follow-up email referenced on record
deposition_02.m4a · t=58:40
Security & trust
Built for evidence you cannot afford to leak, lose, or misstate.
Legal teams, investigators, and enterprises trust us with recorded evidence. Security and provenance are design constraints here, not features bolted on later.
Tenant isolation by construction
Every row carries an organization key and access is authorized on every read. Your evidence is never reachable across tenants — or across other Nimavera products.
Encrypted, served by signed URL
TLS in transit and encryption at rest. Uploads live in private object storage and are only ever served through short-lived signed URLs — never public links.
Two-step verification and revocable sessions
Two-step verification (authenticator app, emailed code or recovery code), required for our platform admins and available to every user; server-side session revocation, and role plus matter-level permissions. The admin console exposes no recordings or transcript text.
Tamper-evident audit trail
Privileged actions land in an append-only, hash-chained log, so the history of who did what cannot be quietly rewritten.
Source hashing & verification
Source hashes, upload manifests, and verification snapshots record provenance and processing history — documenting how material moved through the platform.
Legal hold & expiring shares
Place a matter on legal hold to suspend deletion, and share findings through secure links that expire — access you can revoke at any time.
EU / Spain data support
EU and Spanish data handling with vetted sub-processors under data-processing agreements, listed by role on the sub-processors page.
Malware-scanned, isolated processing
Every file, including attachments and archive contents, is scanned with ClamAV on our own servers before anything opens it; infected files are quarantined. Uploads are validated by signature and codec inside isolated workers, and recovered content is treated as data, never instructions.
What we are honest about
We give you traceable, verifiable material for expert human judgment. We never replace it:
- Tracecanon does not authenticate recordings.
- Tracecanon does not determine admissibility.
- Tracecanon does not produce certified transcripts.
- Tracecanon does not determine truthfulness.
- Tracecanon does not provide legal advice.
Every substantive answer cites its exact source and timestamp, and every conclusion stays traceable to the recording — for a person to review and decide.
AI-generated content may contain errors, omissions or incorrect interpretations. Always verify the result against the cited recording or document.
Honesty as a feature
What it won't claim is why you can trust it.
What lesser tools imply
- “Certified” transcripts
- “Authenticated” recordings
- A verdict on who's telling the truth
- Legal advice by another name
What you actually get
- A machine transcript, honestly labelled
- A byte-exact source fingerprint
- Potential differences, for a human to judge
- A citation — or an honest “not enough”
Every answer cited. Every conclusion traceable.
Trace every answer to what was said. Canon: the authoritative record.
AI-generated content may contain errors, omissions or incorrect interpretations. Always verify the result against the cited recording or document.