Interview, Marta Lindqvist
“The second invoice was the one that didn’t match the contract. I flagged it to my line manager in March and nothing happened. Then in May the auditors arrived.”
Audient transcribes the meetings, voice memos, and talks already on your Mac, then makes the archive searchable in plain English. On-device, no upload.
Not available yet, link goes live when Audient launches
“The second invoice was the one that didn’t match the contract. I flagged it to my line manager in March and nothing happened. Then in May the auditors arrived.”
“Reminder for myself, cross-check the second invoice figure against the filing the company made to Companies House. Marta said £48,000, the filing says £52,400.”
“There were two invoices. The first one was clean. The second invoice went out a fortnight later and that’s the one nobody on the team had signed off.”
“On the record, the second invoice is the document the regulator will want to see first. Don’t lead with the email chain, lead with the invoice.”
Audient is small on purpose. It captures audio and video, turns them into searchable text on your Mac, and lets you cite the exact moment something was said. No accounts, no servers, no syncing.
Ask in plain English and Audient returns the exact passages from across every meeting, call, and voice note in your library. Embeddings are computed and stored on your Mac.
A speech model runs locally and produces word-level timestamps for everything you capture. Audio never leaves the device. Works offline on a plane, in a basement, anywhere.
Every result links back to the exact second in the source recording. Copy a quote and Audient pastes the verbatim line, the file, and the timestamp, ready to drop into a doc or a Slack message.
Audient exposes a local MCP server on your Mac. Pair it with a local LLM like Ollama to keep everything on your machine. Pair it with a hosted one like Claude Desktop if you’d rather not run a model yourself. Audient itself stays put either way.
You decide what enters your library, what gets transcribed, and what gets indexed. Audient does nothing in the background.
Open a file with the picker, drop one onto the window, or share a recording into Audient from anywhere on your Mac. Audient never goes looking for files on its own.
Once you add a recording, the on-device model transcribes it with word-level timestamps. A small badge in the menu bar shows progress.
Hit ⌘K, ask in plain English. Results come back with a direct quote, the source file, and a clickable timestamp that opens the audio at exactly that moment.
Audient exposes its archive over a local HTTP endpoint. Any MCP-compatible LLM client can read it (Ollama, LM Studio, Claude Desktop, and others). Pair with a local LLM to keep everything on your Mac, or a hosted one if you’d rather not run a model yourself. Plain HTTP from your own script works too.
Audient is free to download, with a full trial. A single in-app purchase unlocks it for good. No subscription, no usage tier, no metering.
Available at launch. Localised on the Mac App Store.
Free trial first, then yours forever. No subscription, no account, no cloud.
Drop your email and we’ll send one message when Audient hits the Mac App Store. No newsletter, no drip.