About Whissle
Building personal AI that understands — ethically, sustainably, and in real time.
Whissle began with speech researchers and one stubborn question: how do you capture the full meaning behind spoken words — not just the text, but the emotion, the intent, the context — at the speed of a real conversation?
Where it started
Every existing solution asked you to chain modules together — emotion from one service, entities from another, intent from a third — and most only worked after a sentence finished, at the segment level. Useless for a live conversation, where meaning arrives while someone is still talking.
So the founding team built a different thing: a single discriminative model that extracts emotion, intent, age, gender and entities in one forward pass, running as a modular layer alongside any large language model — richer intelligence in real time, without sacrificing privacy or speed. That model, and the research behind it, is still the core of everything Whissle ships.
The principle underneath it hasn't changed since day one: AI should amplify your capabilities, not replace your judgment — and it should be responsible, accessible, and deeply aware of context.
A dual mission, on purpose
Whissle runs on two engines that pull in the same direction — a non-profit research effort that gives the science away, and a for-profit platform that sustains it.
Non-profit — the science, given away
The research that makes Whissle work is published openly, not locked in a vault. Speech intelligence — understanding emotion, intent and context as someone speaks — is too foundational to be any one company's secret. Our papers, models and methods are meant to be read, cited and built on.
For-profit — the platform that sustains it
The commercial platform — voice agents, the companion, the gateway you can self-host — is how the work pays for itself instead of depending on the next grant. Revenue funds the open research; the open research keeps the product honest and ahead. Two engines, one direction.
A community that grew with the work
What started as a handful of speech researchers is now a global team of researchers, engineers, designers and builders — and a wider community of developers who self-host the gateway, build agents on the platform, and read (and cite) the research. We don't just train models; we build partnerships, and every solution is shaped by real-world needs and grounded in privacy and ethics.
Your data, your choice: self-host on-prem, use the cloud API, or explore the open research — the intelligence is yours either way.