FAQ
Short answers. Where an answer is a limit rather than a capability, it is written that way.
The transcript
What you get back
What happens to audio it could not transcribe?
A span it returned no transcript for is marked on the page, where the words would have been, in every export. The point of the product is that a stretch of missing speech is a visible thing rather than an absence you would have to catch by listening. Audio measured as silence is the exception: it is recorded but not rendered.
The harder case is a stretch that came back nearly empty rather than empty — a transcript entry running a long time on almost no words. There is no absence on the page to mark there, and the completeness figures count those seconds as transcribed, so that case is treated as a finding instead — one that is disclosed rather than blocking. The seconds are reported next to the coverage figure, they are printed on the certificate, and the transcript is delivered. It does not hold the recording: the check reads the shape of the transcript, so it cannot tell a pause that one long entry was drawn over from speech that was dropped.
Will it put a name on a speaker?
Only where the name is supported. A label derived from a call's channel describes a role, not a person, and a speaker the system could not verify carries no proper name at all. That rule is enforced in the database rather than by convention.
Does the coverage check mean the transcript is accurate?
No, and it is important not to read it that way. Coverage compares how much time the transcript accounts for against how much voiced energy the source contained. A recording transcribed into fluent nonsense at the correct timestamps would pass it. It is a completeness tripwire, not a quality measure.
Why was my recording flagged when it sounds fine?
The source-side measurement detects voiced energy and cannot distinguish speech from an automated prompt, hold music, or touch tones. A recording with several minutes of hold music the transcriber correctly ignored shows up as energy the transcript does not account for.
Can I re-run it later and get the same transcript?
Not reliably, and nothing here will claim you can. The transcription service exposes no version pinning. Reproducibility comes from the stored artifact and its digest record, not from re-calling the model.
Your audio
Formats and preparation
Do I need to convert my file first?
No, and it is better not to. Converting is a lossy step on audio that usually has very little detail to spare, and it happens before anything has had a chance to use what was there. Upload the file exactly as it arrived and let the system normalize it once, from the original.
My file is an .amr and will not open. Is it damaged?
Usually not — it is almost always a missing decoder. A player with no AMR support often reports a codec it cannot read as a broken file, because from its point of view the two look identical. See how to play an AMR file.
What happens if the file cannot be read at all?
It fails at the probe with an operator error code, before any transcription is attempted. A loud failure on the whole file is safer than a partial transcript of the portion that happened to decode, because a partial transcript looks complete.
Will heavily compressed phone audio transcribe well?
Expect it to be harder. Narrowband telephony codecs discard information a transcriber would have used, and it is not there to recover. What matters is that the transcript shows you where it struggled instead of smoothing over it.
Handling
Storage, deletion, and Clio
The Clio connection is not available yet — no deployment has it configured, so no firm can connect one today. The answers below describe how the integration behaves once it is.
Who can read my transcripts?
Transcript content is stored as ciphertext under a key held per matter. Backups are ciphertext for the same reason. Access within your firm follows your firm's own matter permissions.
What does deleting a matter actually do?
It destroys that matter's encryption key, which renders the stored content unreadable including in backups that already exist, and issues a record of what was destroyed. It is not a flag on a row that hides it from a list.
Does it write anything to my Clio account?
No. The only write it makes to Clio is the token exchange that establishes the connection. Every product operation — searching matters, browsing documents, fetching a file — is a read, and it runs under your own permissions. See what it does with Clio.
Are my Clio matters copied into Pincite Audio?
No. Matter and document search is live against Clio at the moment you search. There is no local mirror to fall out of date and no background job keeping two copies in step.
The product
Status and scope
Can I buy this today?
Yes. Pincite Audio is in an open beta: create an account and the free minutes come first, with no card collected at sign-up. If you would rather we contacted you, join the waitlist instead. The rate is published — see how pricing works — because a price you cannot find is not a kindness.
Can I use this inside Claude?
Yes, and it is the point rather than an add-on. Claude does not accept audio files, so a connector is the only way a recording's contents reach it. Pincite Audio is added in claude.ai as a custom connector; it is read-only, it reaches only your firm's matters, and every passage it returns carries a citation the server generated. See what the system does.
Is this case management software?
No. It does one thing — turn audio into a citable transcript — and connects narrowly to the case management you already run. It does not want to become the place your matters live.
Why is there no cancel button?
Because the upstream transcription operation cannot be cancelled. A button that claimed otherwise would be a lie about the state of your material, so jobs run to a terminal state and report what happened.
Something here does not answer my question.
The resources section goes deeper on formats and on how the system behaves, the transcript format is the line-by-line reference, and support is where to go if neither covers it.