AI Hearing Transcript

No. AI Hearing Transcript produces a working draft for staff, counsel, or a transcriptionist to review. It does not replace a reporter, a certified transcript, or the record a tribunal keeps. Quotations taken from the draft should be checked against the recording before they appear in minutes, a decision, or a published document.

With speaker separation enabled, each voice is returned as its own labeled turn, so questions from the bench, answers from a witness, and statements from a representative stay apart. Labels can be renamed in the editor to a chair, a member, an applicant, or a witness, which makes a long session far easier to read.

A timestamp turns a quote into a locatable moment. Instead of relistening to a session to confirm that a condition was imposed at a particular point, a reviewer jumps to the time and checks the exact words. The same reference can be cited in a correction request or attached to an appeal submission.

Loud rooms, applause, and a speaker who turns away from the microphone all reduce accuracy. A hearing recorded through a room system usually fares better than a phone on a table. Passages that are unclear should be marked for human review rather than quoted, especially when the wording carries a legal consequence.

Yes, provided the microphone picks it up and speaker separation is on. Public comment often involves many unfamiliar voices, so labels will be generic and may need renaming if particular speakers matter. A comment that is inaudible in the source will also be inaudible in the draft.

Speech up to three hours per file and 100MB per upload are supported. A full-day hearing therefore exceeds the single-file limit and should be split, usually at a recess or after a witness finishes. Splitting also lets staff begin reviewing the morning while the afternoon is still processing.

The service sends audio to the speech recognition provider identified in the privacy notice and does not keep the uploaded audio after the job completes. A body that holds closed sessions should confirm that this processing satisfies its own confidentiality rules, and may transcribe only the portions on the public record.

The text can be corrected in the editor, then exported as plain text for minutes, a briefing note, or a case file, or as SRT when a recording will be published with captions. Speaker labels and timestamps remain in the export, so a reader can still see who spoke and when the exchange happened.