Upload an archive, get the themes running through it

Upload files, or connect a podcast RSS feed directly. Pro takes up to ten hours of audio; Super Pro up to fifty.

memoir. Forty hours of family interviews is not a book and cannot be read linearly. The analysis reads across all of it and reports what the material is actually about — which is almost never the chronology you recorded.

business. A decade of talks, podcast appearances and internal presentations contains your argument, stated dozens of times and never written down. The analysis reports what you have actually been claiming — which is frequently not what you would say if asked — and where each claim is put best.

childrens. Stories told aloud to children — bedtime inventions, the ones a class asks for again — are a real archive and an unusual one. The analysis finds the recurring characters and situations across recordings, which is how a set of improvised stories becomes a book or a series rather than a folder.

fiction. Some novelists talk their way through scenes before writing them, and end up with hours of thinking-aloud alongside a partial draft. The analysis reads across it and reports what the story is actually about — which often differs from the outline, and the recordings are usually right.

graphic novel. Worldbuilding conversations, artist calls and voice notes about panels accumulate fast and become unsearchable. The analysis reads across them and reports the world's actual rules as you have stated them — including the places where you have contradicted yourself, which an artist will otherwise find.

non fiction. Interviews, lectures, seminars and field recordings are the raw material of a great deal of non-fiction, and the difficulty is never gathering it. The analysis reads across everything at once and reports the themes, so the structure comes from the material rather than from your assumptions about it.

poetry. Poets who draft aloud, or who have years of readings recorded, hold a body of material with no index. The analysis reports the recurring images, preoccupations and forms across it — which is one way to discover that the collection you have been assembling is not the one the material is actually about.

script. Improvised scenes, table reads and recorded workshops contain dialogue that is usually better than anything typed cold, because it was spoken. The analysis reads across the recordings and reports what is there — which scenes exist, which characters have a voice, and where the story keeps returning.

textbook. Years of lectures, seminars and tutorials are the best possible source for a textbook, because the explanations have been tested on students who did not understand them the first time. The analysis reports the recurring explanations and where each is put most clearly.

travel. Field recordings, voice notes from the road and interviews with people you met are the most perishable and most valuable travel material — and the least usable in raw form. The analysis reports the themes across all of it, which is how a journey becomes a book rather than an itinerary.

Built for archives. Upload audio files or connect a podcast RSS feed. If your archive is written rather than spoken — years of blog posts, newsletters, sermon notes, teaching material — upload those instead and the same analysis runs across the text.

If you have ninety podcast episodes, or a decade of sermons, or forty hours of interviews with your father, you have already done the hardest part of writing a book. The material exists. What you do not have is any idea what shape it makes, and so you sit down to 'start writing' as though nothing had happened yet.

That is the wrong first move. The first move is to find out what you have already said — which themes keep returning, which story you have told nine times in slightly different words, where the genuine spine is. Almost always the book is already there in the archive and needs assembling, not writing.

It finds what recurs. It cannot tell you which theme you care most about, and that judgement is the book. Expect to override it at least once — the theme you argue with is usually the one worth writing.