Questions

Because a language model regenerates text, it does not copy it. Over four thousand messages it will paraphrase, trim, quietly fix a typo. A reworded message is not evidence any more. ChatExhibits parses deterministically: the same input produces the same output, to the character. It also cannot attest to any chain of integrity, and sending a client’s conversation to a third party is a problem an attorney does not need.

The ZIP is read in a stream and hashed in 8 MB blocks rather than loaded whole, which is what makes several hundred megabytes workable in a browser tab. Very large exports are slower, never rejected, and the progress you see is real rather than a spinner.

A speech model runs on your own machine, in a background thread, and never sends the audio anywhere. It needs a desktop browser. You then read and correct every transcript before the PDF is generated. That step is mandatory, and each block is marked as reviewed or not reviewed, so the reader always knows which is which.

The ZIP that WhatsApp produces when you export a chat, or the .txt on its own if you exported without media. It is read in your browser either way, and the export guide covers getting it off the phone on iPhone and Android.

The document does. Transcribing voice notes does not: it needs a desktop browser, because the archive is already in memory and a phone kills the tab once the speech model joins it. That is not a limitation anyone meets in practice, since the export leaves the phone and the preparation happens at a desk.

The model is multilingual, and you choose the language of the conversation before it starts: English, Spanish, French, Portuguese, German, Italian, Arabic and Hindi. Choosing matters, because left to guess on a short voice note it guesses badly.

No, and nothing can. ChatExhibits produces a document; a judge decides what a court accepts. Bates numbering and a digest make that document citable and checkable, which is a professional convention rather than a legal requirement.

They are embedded as numbered figures in an appendix and referenced from the message they came from, so the page count stays honest and each figure can be cited. Videos carry their first frame, filename and duration.

On the Legal tier you can redact passages and filter by participant or by date range. Redacted text never enters the PDF at all, rather than being covered with a black box that leaves the words underneath. A redaction is visible as [REDACTED] and the message keeps its paragraph number, because a document that shortens a sentence in silence is worse than one that admits the cut.

You are asked, before anything is generated. 3/4/24 is either 3 April or 4 March depending on the phone that wrote it, and when the file itself does not settle it nothing here guesses. The certificate then records whether the order was deduced from the file or stated by a person.

WhatsApp writes every timestamp in the local time of the phone that exported the chat, and never records which zone that was. So you are asked, and the certificate says which zone the times are expressed in. On a matter that turns on a pickup time, a silent assumption would be the whole problem.

Under 10 MB, yes: the archive is embedded so a recipient can recompute the digest from the document itself. Above it, the archive is delivered beside the PDF instead, because filing systems cap attachment sizes, and the certificate says which of the two it is rather than announcing an attachment that is not there.

Nothing. There is no server to keep it on. Your preferences stay in your own browser: the Bates prefix, the last number used, the layout you chose.

Nothing on any server, because there is none. Four small things stay in your own browser: what you paid for, your matters and their numbering, your letterhead, and where your next production should start. No message, media file or transcript is ever written to any of them.

No. We never receive it, so there is nothing to look at and nothing to ask you to send. If something goes wrong, what helps is the error you saw and what the file looked like, never the file.

No. An export is paid for once, and you can regenerate it as often as you like for thirty days: correcting a transcript, changing the Bates prefix, redacting something you missed, switching layout. The price is for the export, not for each attempt at it.

No message limit and no page limit. Four thousand messages or forty thousand, the price is the same. Converters that step their pricing at 5,000 and 25,000 messages are pricing the volume; we price the use, which is the whole difference between a document and a commodity.

The free preview is the trial. It parses your actual export and shows you the result before any payment, so if something goes wrong with your file you find out before you have paid rather than after. That is also why there is no refund policy to read.

It is arranged by hand for now. You request access, we issue it, and exhibits are organised by matter with numbering that continues per matter and a certificate on your letterhead. A subscription that unlocked itself in one browser could not follow a partner to a second machine, could not reach the paralegal who prepares the exhibits, and could never be revoked when it ended.

Sales are final, and the reason is that there is nothing to buy blind: the preview is free, unlimited and needs no account, and it lays out your own conversation exactly as the paid document will look. If a fault on our side stops a document being produced from an export the preview accepted, write to us and we will fix it or refund you.