Making Text to Speech Usable: Introducing Valeon TTS Studio
I’ve been living inside a text-to-speech workflow for a while now—not as a novelty, but as a practical necessity. Valeon has been growing into a publishing system that wants to exist in more than one form at once: writing you can read slowly, audio you can carry with you while walking, and a repeatable process that doesn’t turn creation into a second job.
For a long time, the TTS side of that world was just scripts, folders, and a trail of generated artefacts. It worked because I built it, because I understood it, and because I knew where everything lived. But that kind of setup starts to buckle the moment other people get involved. “How does it work?” becomes a real question. “Can I run it for my own writing?” turns into friction. “Can I try it without cloning a repo and learning your exact structure?” becomes the difference between a tool and a curiosity.
“Clone this repo and run these commands” is fine for people who enjoy that kind of friction, but it’s still a dead-end for everyone else. And even for developers, it’s often a barrier disguised as a solution. I wanted something that takes what already works and makes it usable in the ordinary sense of the word—something you can open, understand, and rely on without needing to inherit my habits just to make it run. That’s what Valeon TTS Studio is: a straightforward interface that turns a working pipeline into something you can actually use day-to-day, without ceremony and without guesswork.
Valeon TTS Studio is deliberately modest in what it claims to be. It isn’t trying to reinvent speech synthesis, or sell you a platform, or wrap your work in a new kind of dependency. It’s trying to make a powerful capability feel usable, trustworthy, and boring in the best way—the kind of tool you can return to without having to think about it too much.
At its core, the app is a wrapper around the OpenAI TTS API, built with a bring-your-own-key approach. That means your usage stays yours, your billing and limits stay in your control, and the key stays in your hands. The design is local-first by intention as well: everything happens on your machine except the one thing that cannot happen locally, which is the TTS request itself. That request must be sent to the remote model to generate audio, and that constraint isn’t a flaw—it’s simply how hosted inference works. The real goal is to make sure everything else doesn’t leak outward unnecessarily.
There’s a deeper reason I care about local-first beyond the usual checklist of privacy and security. Writing—especially writing that’s honest—passes through messy, private, unfinished stages. It’s half-sentences, paragraphs you’ll delete, contradictions you haven’t resolved yet, and scaffolding you’d never want published. A TTS workflow that nudges you toward uploading your entire process into somebody else’s ecosystem can push you into self-censorship long before you realise you’re doing it.
In a time where “convenience” often means “silent surveillance,” I wanted something that feels more like a workshop than a platform. You bring your own tools, you do your work, and you leave with what you created. The API key model is part of that, but so is the emphasis on keeping the experience lightweight, transparent, and auditable. The principle is simple: your drafts should feel like they stay yours, even while you’re turning them into audio.
If you just want to try it, you can use the public hosted studio at Valeon TTS Studio. But the hosted version isn’t the “real” product in the usual SaaS sense. The real product is the code, the approach, and the permission to adapt it. Valeon has always been about building in the open—tools, ideas, pipelines, essays—because the internet is at its best when it becomes a library of working systems rather than a museum of locked doors. If you want to self-host, audit, extend, or repurpose the studio, you’ll find the source code on GitHub at the Valeon TTS Studio repository. My hope is that it becomes useful beyond my own publishing—for anyone trying to make their writing more accessible, for anyone building an audio layer on top of text, and for anyone who wants a clean, honest interface to TTS without the usual baggage.
