TasksBatchTools
Pricing
English
☰

Turn an interview into
an article draft you'll actually edit, not rewrite.

audien·to is an AI audio tool that turns an interview recording into a structured article draft — lede, narrative arc, verbatim quotes preserving the speaker's voice — in under a minute. Unlike auto-summarizers that flatten everything, it leaves the scaffolding for you to edit, not rewrite. Free tier, 67 languages, no signup, audio auto-deleted in 72 hours.

Drop the recording in. We'll lay out a lede, build a narrative arc, and pull verbatim quotes that read like the actual person said them — because they did. You finish the work; we handle the scaffolding.

● Record
Drop audio here
or click to choose a file · up to 2h · auto-deleted in 72h
LANGUAGE
required · 67 supported
What you'll get
  • Title and lede
  • Narrative body
  • Verbatim pull quotes
  • Sources and timestamps
Speakers auto-tagged. Tweak any section after upload from the Options panel — no setup needed up front.
The guide

Why interview-to-article is hard to do well

An interview transcript and an article are two different artifacts. The transcript is faithful but unreadable: people talk in fragments, double back, and qualify everything. The article is readable but loses the voice if you're not careful.

The job is to keep what made the interview worth doing — the actual things the person said — while giving a reader the shape they expect. Lede that hooks, narrative that progresses, quotes that earn their place. We draft that scaffolding; you bring the editorial judgment.

What a good interview-driven article contains

  • A lede with stakes — the first paragraph names the question and why it matters now.
  • A narrative arc — context → tension → resolution. Not a Q&A transcript with line breaks.
  • Verbatim quotes — if you'd put it in quote marks, the words have to be exactly what was said.
  • Voice preserved — the article should still read like *that person* talking, not the publication's house style.
  • Source-able claims — every factual statement traces back to a timestamp in the recording.

How audien·to drafts yours

What lands in the doc

  • Title and ledeheadline + 80-word opening graf that names the question and the answer.
  • Narrative body3–5 sections, each anchored by a claim from the interview. Speaker tagged, paragraphed, ready to edit.
  • Pull quotes3–6 verbatim lines, attributed to the subject, each long enough to stand alone.
  • Source mapevery quote and claim mapped back to its timestamp in the recording, so you can verify.
  • Open questionsany threads the interview left unresolved — surfaces them as a “follow-ups” list, not buried in the body.

Writing from interview audio

  • Quotes are sacredif it's in quote marks, the words must be exactly what was said. Cleanup happens around quotes, never inside them.
  • Lede earns the readopen with the unexpected thing the subject said — not their bio. Bio belongs in graf 3.
  • One claim per sectioneach section answers one question. If a section needs subheads, it's actually two sections.
  • Voice over polishif the subject says “gnarly” — keep it. Smoothing into corporate prose is the failure mode.
  • Source-mapping disciplineevery factual claim should be checkable in the audio. We attach timestamps so you can verify before publishing.

Knobs in the Options panel

  • LengthShort (~600 words) · Standard (~1200) · Long (~2000+).
  • VoiceMatch the speaker (default) · Match the publication · Neutral.
  • Quote count3 / 6 / 10 — fewer is more punchy, more is better source coverage.
  • Lede styleIn-the-room scene · Provocative claim · Stakes-first · Question.
  • Output language23 supported. Translate to publish in any of them.
Why this works

Why modern AI hears what older tools missed

When you upload audio, two AIs go to work. The first one listens. It learned from millions of hours of real speech — accented, overlapping, full of “ums” and brand names that didn’t exist five years ago — so it can hear “Klaviyo,” “Substack,” or “the GPT pipeline” without flinching. The words older tools used to silently mangle come back right.

It hears in context. Instead of guessing one sound at a time, it takes in the whole sentence and uses everything around a tricky word to figure out what was actually said. That’s how a brand-new product name still lands correctly: the words around it tell the AI what kind of sentence it’s in.

And it cleans as it goes. Disfluency — “uh, like, I think… yeah” — doesn’t drop the rest of the sentence on the floor. Punctuation and capitalization come built in, so what you read is prose, not a wall of lowercase. By the time it hands off, the transcript already looks like what a careful typist would have given you.

What happens once we have your words

A raw transcript is the floor, not the ceiling. The second AI reads the whole document the way a careful editor would. It groups related discussion into chapters even when nobody says “moving on.” It surfaces the quote you’d actually screenshot — not the longest sentence on the page. It separates a decision from a tangent, an action item from a passing wish.

That’s the jump older tools couldn’t make: they gave you words, we give you shape. A meeting becomes minutes with owners, dates, and resolved questions. A podcast becomes show notes whose chapters track the real narrative, not the nearest five-minute mark. A voice memo becomes a send-ready email in your voice — not a list of fragments to stitch back together.

Each tool on this page is one of those pairings — the same listening AI up front, the same writing AI behind it, shaped for one specific output. You don’t pick. You don’t tweak. A thirty-second upload comes back as the thing you actually wanted, ready to use.