An index into the testimony, not a shorter transcript
What a summary is actually for is finding testimony when a witness says something at trial. That means citations, and it means covering every volume.
Nothing to upload. We run it on a transcript we supply.
- 01 Input
Multi-volume transcripts, as one document
Testimony spread across several sessions is the case where manual summarising breaks down worst, because a topic index has to span the volumes rather than exist once per volume. A subject raised early, revisited in the middle, and returned to at the end has to be grouped, and that can only be done by something tracking it across the whole record.
Transcripts arrive in whatever form the case produced them: text-based PDFs, scanned or image-only PDFs, photographed and faxed copies, DOCX, plain text, and condensed minuscript printed several pages to a sheet. OCR reads the pages that are images rather than text, but OCR is not the point. Every input, clean text included, is then read by a language model to recover the structure of the transcript — the reporter's page boundaries, the line numbers, the speaker labels, the question and answer pairs, and the exhibit references.
That recovery is what the citations rest on. A citation is only useful if it points to the reporter's page and line, and the reporter's pagination is not the PDF's page count. Court reporter output varies by reporter and by format, line numbers sit in the margin rather than in the body of the text, and a condensed transcript places several reporter pages on one physical sheet. Recovering the reporter's own pagination is what makes the citations in 03 resolvable.
- ✓Text PDFs, scans, photographs, faxes, DOCX, plain text, and condensed transcripts
- ✓Reporter page and line numbers are recovered, not the PDF's page count
- 02 Structure
Segmented by topic, not by page order
Testimony is not organised, and a summary that follows the transcript sequentially reproduces its disorder. The output is categorised by subject, so related testimony sits together regardless of where in the record it was given.
- 03 Cite
Every entry carries a page and line range
This is the axis on which a summary is useful or useless. A summary describing testimony without pointing to it is orientation — you can read it to remember roughly what was said, but the moment anyone asks where, you are back in the transcript hunting. A summary with page and line references is an index into the record, checkable by opposing counsel and by the court, which is what makes it safe to rely on.
- ✓Entries cite page and line range, not page alone
- ✓Citations resolve into the transcript you supplied
- 04 Review
Read an index instead of building one
Outsourced summary services publish their throughput, and it clusters around twenty to twenty-five transcript pages an hour, with standard turnaround around five business days. A multi-volume deposition is therefore days of someone else's attention before you see anything. Processing here runs in the background, and your time goes into reviewing an index rather than constructing one.
See it run on a transcript like the one you're working with. Nothing to upload. Request a demo
What it doesn't do
Stated plainly, so you can decide with the whole picture rather than finding out mid-case.
Review is not optional
Expect to spend real time reading and correcting the output, and budget for it rather than assuming a finished product. The saving is in not building the index, not in skipping the read.
Check citations on a transcript you know first
Citation accuracy is the metric that decides whether this is useful, so test it yourself before relying on it. Run it over a deposition you have already worked, then check whether the references resolve to the right page and line. That takes an afternoon and tells you more than any demonstration.
It indexes testimony, it does not assess it
Which testimony is damaging, which is impeachable, and what to do about it are legal judgments. The summarizer organises the record so you can reach them faster.
Citation accuracy follows input quality
A clean text transcript gives the parser the most to work with. A poor scan, a skewed photograph, or a copy covered in handwritten annotation gives it less, and the page and line references are the first thing to degrade. It is a further reason to check citations on a transcript you already know, and to run that check on material in the condition yours actually arrives in.
Common questions
- What does the output actually look like?
- A topic-categorized index of the deposition. Each entry states the subject and cites the transcript by page and line range, so you can go straight from the index to the testimony.
- Does it handle depositions that span several volumes?
- Yes, and that is the case it is built for. A topic index across multiple volumes is where manual summarising costs the most, because the effort is more than additive — one index has to span every session.
- Can I use the summary to prepare for cross-examination?
- That is what the citations are for. An entry you can trace to a page and line is usable for impeachment; one you cannot is only orientation. Verify the specific citations you intend to rely on.
- How does this compare to an outsourced summary service?
- Human summarisers work at roughly twenty to twenty-five transcript pages an hour with turnaround typically around five business days. The difference is less about speed than about when you can start: you review an index rather than wait for one to be built.
See it on your kind of material
Tell us what you're working with and we'll show you it running on something like it. Nothing to upload.