Narralog

[buy product=”narralog”]

Narralog

record your voice and export a file that meets the spec.

Narralog records a chapter, cleans it, masters it, and then checks the finished file against the loudness, true-peak and noise-floor numbers an audiobook or podcast platform grades on. it runs on your own computer, offline, with no account and nothing uploaded.

Narralog is a recording and mastering tool, not a review service. it masters to the numbers a platform publishes and shows every measurement next to the requirement, so you can see where a file stands before you send it. whether a submission is accepted is that platform’s decision, and each platform sets and revises its own requirements.

what it does

record, or bring a take you already have

capture from the microphone at the device’s own sample rate with automatic gain, echo cancellation and noise suppression switched off, so the voice arrives untouched. or import a file you recorded elsewhere — WAV, MP3, M4A, AAC, FLAC, OGG, Opus, WebM or MP4.

a waveform you can edit

zoom in, drag a selection, play it back to check it, then trim the take down to the selection or cut the selection out. one-key undo. flubs and false starts come out before anything else runs.

measured the way the standard says to measure

RMS, sample peak, noise floor, integrated LUFS to BS.1770, and true peak taken the way that standard says to take it — four-times oversampled with a windowed-sinc reconstruction of the inter-sample values, rather than a polynomial approximation of them.

noise reduction that leaves the voice alone

a second-order high-pass at 85 Hz takes out rumble, handling thumps and the body of mains hum. then spectral subtraction learns your room’s noise print from the quietest tenth of your own take and pulls that hiss down under the voice. it is one slider, from off to strong, sitting at medium out of the box.

dead air out, real pauses kept

out of the box a silence up to two seconds is left exactly as you recorded it, because sentence beats and paragraph breaks are how narration sounds like a person. only a longer gap — a stumble, a page turn, a think — is shortened, and it is shortened to about eight tenths of a second, trimmed evenly from both ends so the cut is not all at the front. both numbers are sliders. words are never joined.

a noise floor that gets fixed

if the room floor sits above the limit, a downward expander pulls the quiet stretches under it while the voice passes at full gain. the pauses keep a low room tone rather than dropping to digital silence, which is what the audiobook spec asks for.

three output profiles

an audiobook profile at -20 dBFS RMS, a podcast profile at -14 LUFS integrated, and a universal profile that sits at the loud edge of the audiobook window with enough headroom left that one file satisfies both. the kind of project you made picks one for you, and you can change it. the whole chain retunes to whichever is set.

room tone from your own room

head and tail padding is cut from the quietest four tenths of a second of your own recording and looped, so the silence at the top of the file is the same silence as the rest of it. three quarters of a second at the head, two seconds at the tail.

the submission check

after mastering, every measured value is set beside the value it has to hit and the file gets a verdict. you watch it pass on your own machine before you upload it, and if something is out you see which number and by how much.

true-peak limiting with a safety pass

a look-ahead limiter holds the ceiling, then the true peak is measured again on the limited signal and the whole file is trimmed down if an inter-sample overshoot survived. the ceiling is verified on the signal that actually goes out, after the stage meant to enforce it.

projects, and what lives inside one

a project is an audiobook, a podcast, a voice diary, or whatever else you are making, and the wording, the buttons and the default profile follow that choice. an audiobook holds chapters and gets its own opening and closing credit slots; a podcast holds episodes, a voice diary holds entries, anything else holds parts. all of them renameable and deletable, with a Trash you can pull something back out of.

export

192 kbps constant-bitrate mono MP3 at 44.1 kHz, which is the delivery format both kinds of platform take, or WAV if you want to carry the master somewhere else.

who it’s for

narrating a book that isn’t yours

you are under contract on someone else’s manuscript and the raw audio is a book nobody has read yet. it stays on the machine it was recorded on, because there is nowhere else for it to go.

a podcast made at a desk

you want the episode to land at the loudness the platform normalizes to, without first having to learn what an integrated LUFS reading is or which meter to believe.

a first audiobook

you have been sent a page of requirements with four numbers on it. the submission check puts your file’s four numbers next to them and tells you which one is out before a reviewer does.

what sets it apart

measured, not estimated
true peak by the oversampling method the standard specifies, and integrated LUFS by the same standard. the reading is taken by the method the spec names, not by an approximation of it.
it repairs the failure
a noise floor over the limit is expanded down. a true peak over the ceiling is trimmed. the measurement is shown before and after.
timing survives
the de-pauser only touches a silence longer than the threshold you set, and shortens it to a paragraph beat. your pacing is still your pacing.
offline
no account, no sync, no server. an unreleased recording never leaves your drive.
one purchase
not on sale yet. when it is: one payment for the desktop app, and updates to the version you bought will cost nothing extra. it runs on Windows today, with macOS and Linux to follow on the same purchase. the phone app is its own purchase.

the numbers it targets

the profiles are built from the published delivery requirements, so the checking happens inside the app.

audiobook
RMS between -23 and -18 dBFS, true peak at or under -3 dBTP, noise floor at or under -60 dBFS, room tone at the head and tail, mono 192 kbps MP3 at 44.1 kHz. these are the ACX requirements for Audible.
podcast
-14 LUFS integrated with a -1 dBTP ceiling, which is the level Spotify normalizes to.
universal
one file that clears the audiobook gates and still holds enough headroom to be normalized down to the podcast target without clipping.
the verdict
pass or fail per number, on your own machine, before anything is uploaded anywhere.

Cosmicore LLC is not affiliated with or endorsed by ACX, Audible or Spotify, and those names are their owners’ trademarks. they appear here to say what the output is measured against. each platform sets and revises its own requirements; Narralog reports what your file measures, and it does not speak for anyone’s review process.

what stays on your computer

Narralog works offline. there is no account to make, nothing to sign into, and no connection needed to record, process or export. your takes, your projects and your finished files are files on your own drive that you can copy, move or delete yourself.

that matters more here than in most apps. a narrator working to a contract is holding hours of a book that has not been published, and a podcaster has an episode that is not out yet. neither one goes anywhere.

and there are no Narralog servers. an app can only hand over, leak or lose what it has — this one has nothing. the privacy page says the same thing in full.

where it is

engine
finished. measurement, denoise, de-pause, mastering and MP3 export all run against a passing test suite.
desktop
0.3.3, running on Windows. macOS and Linux builds are next.
phone build
1.1.6, in progress for iPhone and Android.
buying it
not on sale yet. when it is, the desktop app is one payment covering Windows, macOS and Linux, and updates to the version you bought cost nothing extra. Paddle takes the payment as merchant of record, which makes Paddle the seller: it collects and remits the sales tax and VAT, and it owns the refund relationship — the terms are on the refunds page. there is no licence key and nothing to activate: the download link is on the page you land on after paying, which is worth bookmarking.
phone price
a separate one-time purchase in each store when it ships. the price is not set yet. no subscription, no in-app purchases and no ads on either.

questions go to hello@cosmicore.net.

what you need to run it

system requirements

windows
windows 10 or 11, 64-bit. about 400 MB free. the installer is not code-signed yet, so windows shows an “unknown publisher” box the first time — more info, then run anyway.
microphone
a microphone, and somewhere quiet to use it. anything from a headset up will do; the room matters more than the mic.
internet
only to download it. after that it never needs a connection, because there is nothing to connect to.

macOS, Linux and the phone builds are not out yet — the desktop app runs on Windows today. questions go to hello@cosmicore.net.

[appstores product=”narralog”]

FAQ

common questions

how is Narralog sold?

it is not on sale yet. when it is: one payment for the desktop app, covering Windows, macOS and Linux, and updates to the version you bought cost nothing extra. no subscription and no licence key. the phone app is a separate one-time purchase for iPhone and Android. where each build currently stands is set out in “where it is” above.

does it work offline?

yes. recording, denoising, mastering, measuring and exporting all happen on your own machine with no connection. there is no account and nothing is uploaded.

will it get my audiobook accepted?

it masters to the published numbers and shows each measurement next to the requirement, so you can see how the file measures before you send it. acceptance is the platform’s decision and their requirements are theirs to change.

what does it export?

a mono MP3 at 192 kbps and 44.1 kHz, which is the delivery format audiobook and podcast platforms take, or a WAV if you want to keep working on the master elsewhere.

can i use a take i recorded on my phone?

yes. import WAV, MP3, M4A, AAC, FLAC, OGG, Opus, WebM or MP4, and the file is down-mixed to mono and resampled to 44.1 kHz before mastering, so an imported take goes through the same chain as one recorded in the app.

will it make me sound processed?

the noise reduction sits at medium out of the box and the pause trimming leaves anything under two seconds alone, so the pacing you performed is the pacing that comes out. both are sliders, and you can hear the take before and after.

do i need an expensive microphone?

no, and the app says so on the way in: a soft, quiet room, the mic four to six inches away and slightly off to the side, five seconds of silence at the top of every take, one file per chapter. the room matters more than the microphone.

what will it run on?

Windows today. macOS and Linux builds are next, and the one desktop purchase will cover them when they land. a phone app for iPhone and Android is in progress, and it will be sold separately.