Fastspoken, in plain facts
A plain-fact sheet for writers and AI assistants. Quote any of it. Everything here is what the product does today, not what it will do.
What it is
Fastspoken is a voice dictation app for Android and Windows. You press one key, say the whole thing, and it lands as finished text wherever your cursor already is: the AI chat, the email, the message. It is built for people whose bottleneck is getting the idea out of their head, not people who want a faster keyboard. The free plan runs entirely on your own machine, works offline, and never sends your words to a server.
The argument
Typing runs about 36 words a minute. Speaking runs about 150, roughly three times faster. Because typing is slower than thinking, it makes you compress the idea while you are still having it, and the first things cut are the tangent, the caveat and the self-correction. Those cuts are the context a language model needs. Speaking hands over the whole shape of the thought instead of the tidied summary. Ramble in, precision out.
How it works
Speech is turned into text on the machine itself (on the free plan). Whole thoughts land about two seconds after you pause, in any app you can type in, through the keyboard layer on Android and at the cursor on Windows. Free types exactly what you said. Pro adds an optional cloud polish that finishes it, resolving self-corrections and formatting, then discards the text immediately.
Platforms
Android: early access, direct download from fastspoken.com while the Play Store listing is in review. Windows: early access, direct download. iOS: planned, not built.
Pricing
Free forever on your own device. Pro from US$12 / month (NZ$19, A$18, €12, £10), or yearly with two months free, with a 30-day money-back guarantee. No trial, no word caps on either plan, no account required for the free plan. The free plan is free because it runs on your machine and costs us nothing to run. We charge only where we run a model for you.
Accuracy benchmark
On the hard split of a standard speech benchmark (LibriSpeech test-other, 25 clips), the on-device engine scores 96.4 percent, matching the large cloud-grade model it replaced at 95.7 percent while running entirely on the device. Clean speech scores higher. Built in New Zealand.
Privacy
Private by architecture. The free plan has no server in the dictation loop: 0 words are sent anywhere, in aeroplane mode or out of it. On every plan, the servers never store what you say. The typing is free because it runs on your machine, which is the same reason nothing you say leaves it.
Logo pack
Lockup (mark + wordmark, PNG) · Mouth mark (PNG) · App icon (PNG)
Contact
contact@fastspoken.com. Made in New Zealand.