· Court Transcript Platform
Three kinds of record, and why one template cannot hold them
A court hearing, a deposition, and a 911 call are all transcripts, and almost nothing about their pages is the same. What the apparatus actually differs in, and what supporting all three cost us in the schema.
For most of this platform's life there was one shape of document: a court hearing, laid out on some jurisdiction's published rules. The words went in a numbered grid, a caption page went in front, a certification page went behind, and the only real variable was which court you were filing into.
That abstraction broke twice this month, in two different directions. It is worth writing down how, because the differences are not cosmetic and the pattern is the same one anybody building document software eventually walks into: the body of the document was never the hard part.
What actually differs
Every transcript we produce is a fixed-width grid of numbered lines. Below is what is wrapped around that grid, by type:
| Court hearing | Deposition | Evidence recording | |
|---|---|---|---|
| Front matter | Court caption | Title page, then appearances | Exhibit title page |
| Body | Colloquy and Q/A | Q/A examination | Speaker turns |
| Back matter | Certification page | Three certificate pages | Certificate of Transcriber |
| Identifies | The proceeding | The deponent | The recording itself |
| Signed by | The reporter | The officer, and the deponent | The transcriber |
The bottom row is the one that matters. Each of these documents exists to let a different person attest to a different claim, and the page apparatus is just the machinery for getting that attestation onto paper in a form a court will accept.
Court hearings: the jurisdiction is the variable
This is the case the platform started with, and the one where the work is breadth rather than depth. The document is always the same kind of thing. What changes is whose rules govern the printed page.
California Superior runs 28 lines with footer page numbers and a county venue certificate. The federal JCUS format runs 25 lines and wants witness and exhibit indexes. Florida has Rule 2.535, Texas has its Uniform Format Manual, New York has Part 108. Each is a measured ruleset, derived from published standards and, where we could get one, checked against a certified transcript the court actually accepted.
Switching between them touches no words. Formatting is derived at render time, never stored, so choosing a different court reflows the same record onto a different grid and nothing in the transcript itself moves.
Depositions: the correction process is written into the rules
A deposition is the only one of the three whose error-handling is specified by law. FRCP 30(e) gives the deponent 30 days after notice to review the transcript and to list every change, with the reason for each. FRCP 30(f) tells the officer what to certify.
That single fact drags in most of the apparatus:
- A title page naming the deponent, the date and place, and who the deposition was taken on behalf of.
- An appearances page listing counsel for each side, plus anyone else present.
- A Certificate of Deponent, where the witness attests that they read the transcript, subject to changes noted on the attached errata sheet.
- The errata sheet itself: a ruled form with columns for page, line, the change, and the reason, plus a signature block.
- A Certificate of Officer carrying the 30(f) recitals. The oath was administered. Review was or was not requested. The officer is not related to, employed by, or financially interested in any party.
Note what the errata sheet implies. The rules assume the transcript will contain errors and build a procedure for fixing them after the fact. Our position is that this is a floor, not a target: every low-confidence word and every uncertain speaker attribution gets flagged for a human before the document is ever signed, so that the errata sheet is catching genuine disagreements about testimony rather than the transcriber mishearing a name.
What a deponent changes about their own testimony still belongs on the errata sheet. Nothing in the software rewrites sworn testimony on anyone's behalf.
Evidence recordings: there is no courtroom to describe
Then there is the case that broke the abstraction properly.
A 911 call. A jail phone call. A wiretap intercept. A recorded police interview. A voicemail. These get offered as exhibits, and every assumption baked into a court layout is wrong for them. There is no court. There is no caption, because there are no parties to the recording. There is no judge, no reporter who was present, and nobody who can attest to what happened in a room they were not in.
What an exhibit transcript has to establish is the identity of the recording. So the title page describes the artifact instead of the proceeding: recording description, date and time, source or type, approximate length, source file name, exhibit number, identified speakers, who prepared the transcript and when.
The back page is a Certificate of Transcriber, and its claim is narrower than a reporter's certificate by design. It attests that the transcript is a true, accurate, and complete record of that recording, to the best of the transcriber's hearing, skill, and ability. Not that the events happened. Not that the speakers are who anyone says they are. That the tape says what the page says it says.
What this cost in the schema
Supporting depositions turned out to be the interesting engineering problem, and not for the reason we expected. The body grid was already right: a deposition uses the federal 25-line Q/A layout, which we had.
What was wrong was an assumption buried in the profile format. A transcript profile had a caption and a certification. One page in front, one page behind. That holds for every court hearing format we had built, so nothing had ever pushed on it.
A deposition needs two pages in front, title and appearances, and three
behind, deponent certificate and errata sheet and officer certificate, each
starting on a fresh page. So caption and certification both became lists
of pages rather than single pages. That was the last known additive gap in the
profile schema, and closing it is what made the Standard Evidentiary Format a
configuration change rather than another special case.
The lesson is not subtle and we did not invent it: the moment your document model says "a" anything, check whether the law says "a" too.
What is not built
One honest limit, since the format catalog is public anyway. The Standard Evidentiary Format currently ships for audio, and its title page reads TRANSCRIPT OF AUDIO RECORDING. Video exhibits, bodycam and CCTV and recorded video statements, use the same page shape, and the title has to change to say so. That is a title variant rather than a new format, and it is the next thing on the list rather than something you can select today.
Jurisdiction variants of the deposition profile, FRCP 30 against California CCP against Florida, are deliberately not built. The generic profile is sufficient until a real filing proves otherwise, and building speculatively for a court nobody has brought us is how format catalogs rot.
You can see all three side by side, with the full apparatus each one produces, on the transcript types page.