The voice gets personal — your own narrating voice, a marketplace for it, and a credit balance that carries it all

Photo: Unsplash

The weeks before were about what it feels like to listen to a story: gapless playback across every platform, a synchronized read-along where the text lights up exactly where the voice is, a search that finds what you mean even when you mistype. That was the question of experience. This phase asks a different, much more personal question: whose voice is actually telling the story?

Until now, the answer was: a high-quality AI voice we curated. Nice, but foreign. As of this phase, the answer can be: your own. OutaStory can now read a story aloud in the author's own cloned voice — you record a short sample once, and that same voice then narrates every chapter of every one of your own stories. That's the biggest thread of these weeks, and around it a whole little economy sprang up all at once: a credit system with a real wallet that carries these features; a marketplace where you can lend out your voice and narrate with other people's voices; the ability to maintain several audio versions side by side for a single story; and, still hidden for now, a professional cloning tier for studio quality. On top of that, the profile has gotten rounder and the catalog has grown a good bit.

This is the phase where OutaStory stops being a platform where you read and listen to stories and starts becoming one where you contribute something with your own voice — and where, for the first time, real money and real credit flow. That's exactly why an extraordinary amount of care sits underneath the nice features; this post deliberately tells a bit of that story too.

Your own voice tells the story

The core is quickly described and still feels like magic: you deposit a short voice sample once — either directly in the browser via the built-in recorder or as an uploaded file — and OutaStory generates a voice clone from it. From then on, any of your own stories can be narrated in that voice, chapter by chapter, with the same narrative flow the rest of the audio pipeline already delivers.

So this doesn't look like a tech demo but like a rounded feature, there's a dedicated "My Voice" page in the profile. It walks you through the entire lifecycle of a cloned voice:

• Record or upload. A recorder right on the page, with a level meter and a given read-aloud text, so the sample fits in content and length. Anyone who'd rather use an existing file uploads it — both paths end up in the same check.
• Quality is checked in advance. Before anything gets cloned at all, the platform looks at the recording: long enough, clean enough, no gross outliers. That way a clone doesn't fail only after the expensive generation step — you get honest feedback right away if the sample isn't good enough.
• A status that's honest. While the voice is being created, the page shows exactly that — and as soon as it's finished, it jumps to a clear "ready" state with a short sample clip that's cached and plays instantly. So you hear immediately how your own voice sounds as a narrator, without having to narrate an entire story first.
• Re-record if it doesn't fit. If you don't like the clone, you can record it again. More happens in the background than you see — more on that below — but for the author it's simply a button: "again".

The first cloned voice is free. That's deliberate: everyone should get to experience, without a hurdle, what it's like to hear their own story in their own voice. Only once you create additional voices — a calmer one for evening reading, a different one for a different genre — do you pay for it. Which brings us to the next thread, without which this whole area wouldn't work.

Credits — the shared currency

Photo: Unsplash

A cloned voice and an audio version narrated in that voice cost real money — compute time at the provider, money per character for narration. So that this can be billed fairly, transparently, and uniformly across every feature, OutaStory now has a comprehensive credit system with a real wallet per account.

The idea is deliberately simple: instead of pricing every premium feature individually, there's a shared currency — credits — that you top up once and then spend wherever you like. An extra voice, another audio version, future premium features: everything draws from the same pot.

Under the hood, this wallet is built as an append-only ledger — every top-up, every spend, every refund is its own, immutable entry. Nothing gets overwritten after the fact; the balance results from the sum of all entries. That sounds like bookkeeping, but it's exactly what you want with real money: a gapless, traceable trail, where a charge and its refund cleanly belong together and where a duplicate-delivered request never gets charged twice.

You can top up in two ways, and both went live during these weeks:

• On the web via Stripe — the tried-and-tested payment path, secure and familiar.
• Directly in the apps via the stores — on Android via Google Play Billing, on iOS via Apple's in-app purchase. This is the least visible and simultaneously most involved part: an in-app purchase must be genuinely verified server-side before credit is granted — otherwise it could be forged. Exactly this server-side verification of purchase receipts against Google and Apple is now cleanly in place, including the necessary certificate checks, so that only real, store-confirmed purchases generate credit.

For users, pleasantly little of this is visible: you see your balance, a history list of entries, and when setting up a paid feature, a clear statement of what it costs and what's left afterward — no hidden deduction, but a clear notice at the right moment. If the balance isn't enough, there's no half-completed process, just a friendly notice and the path to top up.

Sharing voices — a small marketplace

Photo: Unsplash

As soon as people can clone their own voice, a nice question arises almost by itself: Couldn't I have my story narrated in the voice of someone I like? That's now possible — as a small, voluntary voice marketplace.

Anyone who wants to can make their cloned voice available. Others can then select that voice and have their own story narrated with it — in exchange for credits. The money doesn't disappear into thin air: the voice owner earns a share of every use, and the platform keeps a small portion for generation and operation. What was a purely technical feature thus becomes a small ecosystem in which a distinctive narrating voice is actually worth something.

So nobody buys a pig in a poke, there's a sample clip of the other person's voice before you spend anything: listen first, decide after. And here too the same care applies as with your own wallet — the billing between user, voice owner, and platform is built so that every use is booked exactly once and so that a failed narration cleanly refunds the payment instead of making someone pay for something that never came into being.

It's deliberately a small, quiet marketplace — not a loud bazaar, but an offering for those who like to share their voice, and for those who want to give their story a particular sound. But it's the first place on OutaStory where people give each other something directly — and that feels like more than just another feature.

Several audio versions, one story

A story rarely has only one right sound. The same narrative can sound calm and measured or bright and quick, in your own voice or in a borrowed one. That's why a story can now carry several audio versions side by side.

The first audio version is still included in the normal narration path, as before. Anyone who creates additional versions — a different voice, a different tone — pays for it out of the credit pot. All versions are preserved; nothing overwrites anything else. The author decides which version is the active one — the one listeners get by default — and can remove versions they no longer need.

For listeners, this shows up exactly where it makes sense: if a story has more than one finished audio version, a subtle switcher appears in the player. You choose a version, and playback switches seamlessly to that exact voice — chapter skipping, scrubbing, and synchronized read-along included, since each version brings its own complete segment structure. If a story has only one version, the switcher stays invisible, and everything stays as simple as before.

It's one of those features that looks unassuming and demands surprisingly much in the engine room: every version has to store its own audio files cleanly separated, so a second voice never overwrites the first, and deleting one version never takes another's files with it. Exactly this clean separation — each version in its own, clearly named storage path — is the invisible prerequisite for "several voices for one story" to work reliably. And because credits flow here too, the same iron principle applies: charge first, then generate; if something goes wrong during generation, the amount is refunded — nobody pays for a version that never came into being.

Reliability behind the scenes

A cloned voice isn't just a file — at the provider, it occupies a slot, and there's only a limited number of those. If everyone creates several voices, re-records, and discards them again, someone has to make sure these slots are managed cleanly, that a newly recorded voice reliably replaces the old one, and that nothing is left orphaned. Exactly this lifecycle hardening went in during these weeks.

Three things are part of it, all invisible and all important:

• Slot management. The limited voice slots at the provider are now managed as a shared pool, cleanly assigned and released again — instead of "no slots left" eventually surfacing as a cryptic error.
• Transparent re-cloning. If someone re-records their voice, the old clone is orderly retired in the background and replaced by the new one — for the author it's one button, behind it a clean handover.
• Server-side audio analysis. The quality check of the voice sample now runs reliably on the server, with a real audio analysis instead of rough estimates — so "too short" or "too quiet" is caught before cloning, not after.

This is the kind of work nobody notices as long as it works — and exactly because of that, it's the difference between a nice demo and a feature you can entrust your own face, or rather your own voice, to.

A professional voice clone is waiting in the wings

For anyone for whom the quick clone from a short sample isn't enough, the next tier is already prepared: a professional voice clone in studio quality. It uses a larger sample set and an additional verification step that ensures it's genuinely your own voice being cloned — and delivers noticeably higher fidelity as a result.

This tier is already fully built, but deliberately not yet activated — it runs "dark" behind a switch that will only be flipped once every legal and privacy question has been finally clarified. We're mentioning it here anyway because it shows where things are heading: from "already sounds surprisingly like me" to "sounds like a professional recording of me".

The catalog keeps growing

Photo: Unsplash

Alongside the big voice thread, the catalog kept growing — specifically in an area that had been underrepresented until now: 23 public-domain German children's classics have been added. Stories you read aloud before someone can read for themselves — and thus exactly the kind of text that benefits most from narration and synchronized read-along.

On top of that come the already established pre-generated audio versions and covers for the curated flagship stories: wherever we've generated an audio version once, it's simply carried over as a finished audiobook the next time the platform is built — instantly playable, without regeneration. So the library grows not just in breadth, but also in the share of stories you can listen to right away.

The profile gets rounder

While the voice was the headline, the profile quietly grew up — in exactly the spots where it was rough before:

• First, middle, and last name separated. The name can now be cleanly broken down into its parts and edited; the display is composed from them, instead of — as was the case for some accounts — showing the email address as a fallback.
• Birthday greeting at the right time. Anyone who wants one gets a little email on their birthday — at 9 a.m. in their own time zone, not sometime by server time. There's now a time zone selector in the profile for that.
• Newsletters by the clock that counts. The daily and weekly newsletters also go by the recipient's time zone — the daily one arrives in the morning, the weekly one in the afternoon, each at local time instead of the middle of the night.

These aren't big features, but they're the small courtesies that make the difference between a platform feeling foreign or familiar.

Open cards: an about-us page that shows what's running

A platform handling real credit and cloned voices needs transparency. That's why there's now an about-us/versions page that openly shows which building blocks are currently running in which version — and that credits the many open-source projects OutaStory builds on. Anyone who wants to know what's under the hood finds it there instead of having to guess.

In the same spirit, there's a small but practical tools page for search: the search index can be deliberately rebuilt on demand, so new stories and corrections become reliably findable — the direct continuation of the new search from the previous weeks.

Numbers, for the record

• Pull requests: well over four dozen since the last regular blog post — clearly focused on the voice thread (own voice, credits, marketplace, lifecycle, professional clone, in-app purchase) and the multiple audio versions per story.
• Main topics: author self-narration with a cloned voice; a comprehensive credit system with an append-only ledger; topping up via Stripe and native in-app purchases on Android and iOS; a marketplace for lending voices with fair revenue share; several audio versions per story with a switcher; lifecycle hardening for voice slots; a prepared, still-hidden professional cloning tier; 23 new German children's classics; a rounder profile with names, time zone, and time-zone-aware emails.
• Catalog: still in the thousands, now expanded with German children's classics and with a growing share of instantly playable audio versions.
• Tests: still green in the high tens of thousands and specifically expanded around the new, money-carrying paths — wallet, spend and refund, in-app purchase verification, voice lifecycle, and the multiple audio versions — across unit tests, browser tests, and the end-to-end net against the running platform.

What's next

• Unlock the professional clone — once the legal and privacy clarification is complete, the prepared studio tier can step out of hiding.
• Gently bring the marketplace to life — more convenience for discovering and offering shared voices, without making it loud.
• Try multiple audio versions more broadly — keep fine-tuning the switcher across long stories, different voices, and devices.
• Turn on minor protection — the age engine still waits for our data protection officer to approve the verification method.
• App.Shared Razor Pages tests — the last red entry in test coverage, still ongoing in weekly waves.

This phase shifted something: until now, OutaStory was a place where you receive stories — read, listen, read along. Now it's a place where you put something in with your own voice — and where real credit flows for the first time. Both together demand a different kind of care, and that's exactly what sits under every one of these features: every credit that's charged can be traced back; every spend that fails gets refunded; every voice sits cleanly separated from every other. If anything catches your attention while cloning your voice, topping up, switching between versions, or lending your voice out that doesn't feel right — a sample that doesn't sound like you, an amount that doesn't add up, a switcher that plays the wrong voice —, let me know. This is exactly where voice and money meet, and it's exactly where I want to look especially closely.