Dictave
Blog · Text
What your voice is worth
Speech is the most personal data stream you produce, and the market has noticed. It is worth being precise about what is actually in it.
A transcript of a day's dictation contains your unsent messages, your half-formed opinions, the names of your colleagues, your doctor, your children. That much is obvious. The recording underneath it contains more: your accent and where it is from, your mood, your fatigue, markers of your health that you yourself cannot hear, and the voiceprint itself — a biometric identifier you cannot rotate like a password. Recent research on voice imitation should be read by everyone: a few seconds of audio is now enough to make a convincing copy of a voice. The papers are on our reading list; they are not comfortable reading, and that is the point.
None of this means speech technology is bad. It means the custody of the audio is the entire question. The same capability is a different product depending on where it runs: dictation that processes speech on your own machine holds none of the above for anyone; dictation that processes it elsewhere holds all of it for someone, under a policy you are invited to trust and cannot inspect.
Policies are also not physics. A policy is a promise made by a company at a point in time; companies change owners, change incentives, get breached, and get subpoenaed. The audio that was never collected is the only audio that is reliably safe from every one of those futures. This is why we keep saying the boring sentence: it is not that Dictave chooses not to send your voice anywhere — it is built without the ability to do so.
So: what is your voice worth? To advertisers, a profile. To an impersonator, a key. To you, it is closer to a fingerprint attached to your inner monologue. Price it accordingly, and ask every product that listens to you one question first — where does the audio go? — and accept only one answer you can verify. Ours is written down, in plain language.