Black Friday offers.

See offers
Pepys

The Pepys perspective

Voice-Native Work

Information should be able to change form.

An interview starts as a conversation. A report ends as a document. Somewhere between the two, someone has to replay the audio, find the evidence and write. That conversion is part of the work. It deserves better tools.

Updated November 22, 2026 · Published by Pepys

Mr Pepys writes in an open notebook as sound waves curl around the pages

What is voice-native work?

Voice-native work is a way of creating and using information in which speaking and listening are everyday working methods, alongside reading and typing. Speech can become a document; a document can become something you listen to. The format follows the task.

We use Voice-Native Work to connect established practices: transcription, dictation and text-to-speech. It is Pepys' organizing idea for the work between those tools. The established software categories help you choose a product. Voice-Native Work explains how the jobs connect.

The Keyboard Bottleneck

Keyboard-first work assumes that creating information means typing it and consuming information means looking at a screen. That assumption adds work whenever the source or the person needs another format.

A recording can contain the answer and still be awkward to use. A thought can be clear when spoken and slow to put on a blank page. A saved report can sit unread because reading it requires another hour at a screen. The bottleneck is the effort of moving information into a usable form.

The practical question is where speech or listening removes that effort. Keyboards still earn their place for exact edits, code, tables and quiet shared spaces.

Scroll across the table to compare both approaches.

Two ways to approach the same task
TaskKeyboard-first workVoice-native work
Capture an interviewWrite notes while trying to follow the conversationRecord with permission, then review a timestamped transcript
Write a first draftCompose each sentence through a text fieldSpeak a rough explanation, then edit its structure and wording
Find a source quotationReplay a recording until the passage turns upSearch the transcript and check the matching audio
Read a long documentReserve uninterrupted screen time for every pageListen to suitable passages and read the parts that need inspection

Three ways information changes form

Choose the direction before choosing the app. Recording transcription, live dictation and document listening solve different jobs, even when they share speech technology.

01 · AI transcription software

Recorded speech → usable text

Start with
An interview, podcast, lecture or meeting recording
Work with
A transcript you can search, correct, quote and export

Start here when the words have already been spoken. A transcript makes a long recording easier to navigate; timestamps let you return to the source before using a quotation.

02 · AI dictation software

Live speech → written work

Start with
An idea, explanation or rough draft spoken into your Mac
Work with
Text at your cursor, ready for editing

Start here when you know what you want to say and need to write it down. Dictate the explanation in full, then use the keyboard to tighten sentences, correct names and arrange the argument.

03 · Text-to-speech and document readers

Written information → listening

Start with
A report, PDF, ebook or page of notes
Work with
Read-aloud audio you can follow with the text

Start here when the information is already written. Listening gives you another way through the material. Keep the document available for tables, citations and anything that needs a closer look.

These are separate Pepys products. Transcribe is a web app, Dictate is available for Mac, and Reader is live as a free web app. Choose each for its own job; there is no automatic handoff between them.

What this looks like at work

Reporting: keep the quotation attached to its source

After an interview, transcribe the recording and search for the passage you remember. Check the speaker, wording and context against the audio before quoting it. Dictation can help with the story draft afterward, but a polished paraphrase must never pass for the source's words.

The interview transcription workflow →

Research: separate the participant's words from your interpretation

A researcher can use a transcript for coding and dictate a separate analytic memo after the interview. Keep those documents distinct. The transcript records what was said; the memo records what the researcher thinks it means. Choose a tool your institution permits before uploading participant material.

Compare research transcription software →

Writing: say the explanation before polishing the sentence

For an email or proposal, put the cursor in the draft and explain the point as you would to its recipient. Review the resulting text before sending. Dictation helps with getting words onto the page; checking the commitment, amount and tone is still the writer's job.

Dictation for writing and editing →

Document review: listen until the format stops helping

Read a report's narrative sections aloud, then return to its charts and references on screen. Listening is useful for following an argument and hearing awkward prose. A spreadsheet, equation or dense table usually needs a visual pass.

How to listen to a PDF →

Meetings: retain the discussion behind the summary

When a meeting is recorded with permission, a transcript gives the team a way back to the discussion. Use a summary to orient yourself, then check the recording before assigning an action or attributing a decision. Pepys Transcribe works from the recording you provide; this workflow does not require a Pepys meeting bot.

Meeting notes from a recording →

Which companies build relevant software?

Pepys builds across transcription, dictation and listening. Other companies build tools for parts of this work, under established names such as voice typing, transcription and text-to-speech. The examples below describe those jobs; they do not imply that these companies use or endorse Pepys' category language.

Product examples checked against the linked official pages on November 21, 2026. This is a map of tasks, not a ranking or an accuracy benchmark.

Where the approach helps, and where it stops

A transcript makes a recording searchable. Dictation gives a rough draft somewhere to start. Read-aloud software offers another route through a document. The benefit comes when the new format helps you do the next task with less replaying, retyping or waiting for screen time.

Conversion also introduces errors. Speech recognition can miss a name or a negation. A summary can omit the qualification that matters. A synthetic voice can mispronounce a technical term. Keep the source and check the parts you intend to rely on.

Where processing happens matters too. Pepys Transcribe processes uploaded recordings on servers. Pepys Dictate runs recognition and formatting on your Mac; setup, activation and updates can use the network, and the app you dictate into may sync its text. Pepys Reader processes documents and generates speech on your device in the web app. Choose the product and destination that fit the material.

For sensitive interviews, the first decision is permission and the organization's rules, before convenience. Our research recording checklist explains what to check. Pepys' working paper on university guidance examines those institutional decisions; it does not test or certify products. Accessibility needs may also call for assistive technology beyond these products.

From keyboard-first work to voice-native work

We believe knowledge software should let people change how they work without losing the information they started with. A useful transcript points back to the recording. Dictation leaves an editable draft. Listening keeps the written source available.

That would change how teams judge their tools: whether an interview can become evidence in a report, or a draft can be reviewed aloud, matters more than how much audio an app can collect. Our view is that the handoff between formats deserves as much attention as capture.

Start with one repeated task. Keep the source, use the format that makes the next step easier, and judge the result by the work you finish.

Read the essay: The Keyboard Bottleneck →