Meaning, in context.
We’re designing readings around the surrounding mathematics, with the source notation always in view.
Ideas are made to be understood.
We’re building a voice that speaks the language
of mathematics—with clarity, care, and feeling.
Listen to our first experiment alongside the supplied stock voice. Same words. Original recordings.
An early look at the voice we’re developing.
The paper studies concentration inequalities for continuous-time and discrete-time additive functionals of multivariate diffusion processes. The functions may be unbounded, and the processes need not be reversible.
These are synthetic recordings of mathematical research prose, selected from a small development experiment after founder listening. They are not an equation demo or an independent performance benchmark. Evaria’s experimental candidate received two training updates from nine synthetic examples; the stock recording uses Qwen’s supplied Ryan voice. Ryan is a temporary, nonexclusive voice.
Text adapted from Concentration analysis of multivariate elliptic diffusions, Lukas Trottner, Cathrine Aeckerle-Willems, and Claudia Strauch (2023), JMLR, under CC BY 4.0. Wording adapted for spoken English. Full recording and model provenance ↗
A beautiful voice is only the beginning. Our focus is the relationship between what’s written, what’s meant, and what’s heard.
“The integral from zero to one of x squared, with respect to x, equals one third.”
ILLUSTRATIVE READING · TEXT ONLYWe’re designing readings around the surrounding mathematics, with the source notation always in view.
Pacing and pauses should support the idea. A proof needs a different rhythm from a paragraph.
Visible speech plans, careful listening, and independent evaluation are central to the product we’re building.
We’re starting with mathematics—and with the people who need a better way to hear it. Help us choose what comes next.
Imagine lessons and explainers where the equations feel as natural as the introduction.
For reading fatigue, revisiting a difficult concept, or making mathematical material more accessible.
Prepare the reading. Inspect the script and credit quote. Generate when you’re ready. That’s the workflow we’re building.
Explore the API preview Private engineering preview · Access by invitation// Start with the mathematics.
{
"domain": "mathematics",
"language": "en",
"segments": [
{
"kind": "latex",
"text": "\\int_0^1 x^2\\,dx"
}
]
}Join the early-access list today. Here’s how we’re thinking about paid plans for launch.
PROPOSED PRICING · NOT YET AVAILABLEFor curious people with something to say.
A proposed plan for lessons and explainers.
A proposed plan for products and workflows.
Clear words. Clear usage. Our proposed unit is one credit per character of prepared speech, including spaces and punctuation. Equations expand into words; you’ll see the quote before generation.
USD. Prices, allowances, and features are proposals, subject to measured costs and trial feedback. No purchases or billing are active. Final terms will be confirmed before any payment.
You can hear our first experimental recordings now and request early access. The editor and API are in development. Trial invitations will be phased; joining the list does not guarantee an invitation or a launch date.
Mathematical notation is compact and context dependent. Bounds, subscripts, grouping, and pauses can change what a listener understands. We’re building a workflow that preserves the source, makes the intended reading visible, and lets people review the audio.
No. It is a small engineering experiment based on Qwen3-TTS, using the supplied Ryan voice. The selected clips give an early impression of delivery. They do not establish mathematical accuracy, state-of-the-art performance, or an advantage over other TTS products.
Our first focus is English mathematics with typed text and LaTeX segments. Whole-document PDF ingestion, chemistry, biology, and additional languages are future possibilities, not launch commitments. Tell us what you need in the early-access form.
Our proposed unit is one credit per Unicode character in the prepared spoken script, including spaces and punctuation. Expanded equations may use more characters than their source notation. The intended workflow shows the script and credit quote before generation. Prices and allowances are proposals and may change before launch.
Yes. Joining the list is free and requires no payment details. Any trial limits and any future paid terms will be explained before you choose to participate. There is no paid subscription or checkout on this site.
Help shape a new voice for mathematics.