Voices that sound human — OutaStory receives an ElevenLabs grant

Photo: Jukka Aalho on Unsplash

Some messages land in your inbox and you read them twice just to make sure you understood them correctly. This week was one of those: OutaStory GmbH has been accepted into the ElevenLabs Grants program. ElevenLabs is the company behind — for many — the most natural AI voices currently available, and the grant gives us access to exactly that technology.

For some projects, that would be a nice footnote. For OutaStory, it hits a nerve. Since the very first day of the relaunch, one sentence has stood above everything we build: every story should be not just readable, but listenable too. Audio isn't an add-on for us that might come someday — it's a cornerstone. And the difference between a narration that sounds like a train-station announcement voice and one that feels like someone sitting next to you reading aloud is exactly the difference this grant is about.

This post tells three things: who ElevenLabs actually is, why good voices are so decisive for a platform like OutaStory — and what this grant concretely makes possible, without me promising you a finished feature that doesn't even exist yet.

Who is ElevenLabs?

Photo: Pawel Czerwinski on Unsplash

ElevenLabs is an AI company founded in 2022 that has specialized in a single, hard problem: speech that sounds real. Not "intelligible" — many systems have managed that for years. But real: with emphasis in the right place, with pauses that mean something, with the quiet rise and fall of the voice that separates a sentence being read aloud from a sentence being narrated. Anyone who's heard a classic text-to-speech reader and then a modern ElevenLabs voice hears the leap immediately. It's the difference between "a machine is pronouncing words" and "someone is telling a story".

Technically, there's a whole family of models behind it. The best-known building blocks, simplified:

• Text-to-speech — written text becomes spoken language, at a quality deliberately built for reading aloud, audiobooks, and long-form content, not just short announcements.
• Multilingual support — the same technology works across dozens of languages, including German, which for a platform with a German and English catalog isn't a detail but a prerequisite.
• Voice design — the ability to give a narrating voice character: calmer or livelier, warmer or more matter-of-fact, matched to the story.
• The Agents platform — the newer, exciting branch: not just reading aloud, but listening and responding. Voice-driven, conversational assistants that can hold a conversation.

That last point is also the focus of the Grants program — the accompanying materials revolve heavily around the Agents platform, office hours for grantees, and a developer community. So ElevenLabs isn't just handing over the technology; it's inviting you to build something with it. That's a different stance than "here's a discount code".

Why good voices matter so much for OutaStory

Photo: C D-X on Unsplash

To explain why this news made me so happy, I have to rewind a bit. OutaStory already narrates every published story automatically today. Behind the scenes, a pipeline runs that I wrote about in more detail in an earlier post: an AI model turns the chapter text into cleanly structured, verified SSML — that's the "sheet music" for speech synthesis that determines where to emphasize, where to pause, how to pronounce — and a high-quality speech synthesis turns that into an MP3 that plays in the reader. This pipeline is robust, it runs, and it delivers decent results.

But "decent" isn't the goal. Reading aloud is an intimate thing. Anyone who was read to as a child doesn't remember the words, but the voice — the pace, the small voice changes between characters, the feeling that someone was narrating for them. A platform that offers stories to people of every age can't be satisfied with the audio version being technically correct. It has to feel good. It has to be something you choose voluntarily because listening is nicer, not just more convenient.

And audio at OutaStory is more than a comfort feature. It's a piece of accessibility:

• People with low or no vision read with their ears.
• People with dyslexia often find it much easier to get into a story through listening.
• Children who can't yet read fluently can already be right in the middle of a story.
• And everyone else reads on the go, while cooking, while commuting, with eyes closed before falling asleep.

The more natural the voice, the more of this promise can actually be kept. An artificial-sounding voice is tiring; you put up with it for a few minutes. A natural voice you forget about — and that's exactly when you listen for hours. That's why the quality of narration isn't a cosmetic issue for OutaStory, but a question of how many people our stories can even reach.

This is where ElevenLabs comes in. The grant gives us the opportunity to test exactly this leap: narrations that come closer to what a good human narrator delivers — across every story, automatically, in German and English, without needing a voice actor in a studio for each of the thousands of stories in the catalog.

What the grant concretely means — and what it doesn't

Photo: Sincerely Media on Unsplash

I want to stay honest and sober here, because that's part of this project. A grant is not a finished feature. What we got is access and credits — and with that, the freedom to try out ElevenLabs calmly and seriously, without every minute of speech synthesis immediately hitting the budget of a still-young platform. That's exactly the kind of tailwind a small team needs to dare something ambitious instead of sticking to the obvious.

What this practically means over the coming weeks:

First — evaluate instead of rushing. We'll pit ElevenLabs against our current narration: the same chapters, the same SSML foundation, once as it is now and once with ElevenLabs voices. Only if the difference is noticeably better in everyday use — and not just in a cherry-picked demo sentence — does it make sense to rebuild the pipeline. The existing audio generation stays untouched and stable for now.

Second — finding the right voices. A good narrating voice for a dark fantasy novel isn't the same as one for a light romance or a children's classic. Part of the work is choosing a small, curated set of voices that fits the breadth of our catalog — and that convinces just as much in German as in English.

Third — a look at the Agents platform. This is the part I'm deliberately not making any promises about yet, but that excites me the most. Reading aloud is a monologue. The Agents platform could, down the line, turn that into a dialogue: a reading companion you can ask a question about the story, that reads a passage again, that explains a tricky word for younger listeners. That's explicitly music of the future. But it's the kind of music of the future for which a grant like this sets up the music stand.

The grant comes with a small quid pro quo, and I'm happy to fulfill it: ElevenLabs asks that the Grants logo be shown visibly and linked to the program. That's why you'll now find ElevenLabs on our partners page, with its own logo tile — fitting in both light and dark mode — and a short text about it. It feels right to openly name the companies whose shoulders OutaStory stands on: Microsoft Azure as the foundation, Auth0 for sign-in, our tax advisors in the background — and now ElevenLabs for the voices.

A word on AI voices and responsibility

As excited as I am about the technology, an honest paragraph belongs here too. AI voices are powerful, and power demands care. For OutaStory that means a few clear lines: we use synthetic voices transparently — a narrated story is identifiable as AI-narrated, it doesn't pretend to be a human recording. We use the technology to make stories more accessible, not to secretly imitate the work of voice actors. And we stick with what already applies at OutaStory anyway: content continues to go through the same content moderation, whether read or listened to. A better voice changes none of these principles — it just makes them all the more important.

Numbers, for the record

• Occasion: OutaStory GmbH's acceptance into the ElevenLabs Grants program — access and credits for ElevenLabs' speech and Agents technology.
• Current audio status: every published story is already narrated automatically (SSML pipeline with high-quality speech synthesis); this runs stably and stays unchanged for now.
• Catalog: unchanged and broad — the thousands of stories from previous weeks are live; the grant affects the how of narration, not the what of the catalog.
• Partners page: expanded with ElevenLabs, including a theme-aware grant logo (light/dark) in accordance with the program's terms.

What's next

• A/B comparison of narration — pit the same chapters against each other with today's synthesis and with ElevenLabs voices, and honestly assess whether the difference holds up in everyday use.
• Curate a voice set — a small, carefully chosen set of narrating voices for the breadth of the catalog, German as well as English.
• Explore the Agents platform — investigate the fundamentals of conversational voices and honestly examine whether and where a reading companion truly helps readers.
• Lock in transparency — keep the labeling of AI-narrated content clean and visible as the voices get better.

It's rare for a single email to brighten the direction of an entire area of the project. This one did. Audio has been the heart of OutaStory from the very beginning — the part that turns a story from "text on a screen" into "someone is telling me something". With ElevenLabs behind us, we now have the chance to make that heart beat a good deal more human. I'm excited, and as always, I'll keep you posted. If you listen to a narrated story and something catches your ear, good or bad: let me know. That's exactly what this alpha is for.

— Thimo