About
Your pet has been talking to you this whole time.
You already catch some of it. This is about the rest — and about why we built something that would rather admit it doesn't know.
Anyone who has lived with an animal knows the sounds are not all the same sound.
There is the three-in-the-morning yowl from the hallway. There is the specific bark that means the doorbell, which is nothing like the bark that means a fox in the garden, which is nothing like the bark that means you have been on that laptop for two hours. There is the small rising chirrup that only ever happens in the thirty seconds after you walk through the door.
You know these are different. You know, roughly, what each one is for. And if somebody asked you how you know, you would struggle to say.
That gap — between how much you understand and how little of it you can explain — is where this app lives.
A cat's voice is something it made for you
Here is the fact that changed how we thought about the whole project. Adult cats barely meow at each other. Among themselves they mostly use scent, posture and the angle of a tail, plus a small close-range vocabulary — the hiss, the growl — that is about conflict rather than conversation.
The meow is something cats developed for people. When your cat looks straight at you and makes that sound, it is using a channel it opened specifically to reach a human.
And because that channel is opened between two individuals rather than shared across a species, the researchers who study it describe it as closer to a private arrangement than to a language. Your cat's feed me is not the cat next door's feed me. It is a small dialect with two speakers in it, and you are one of them.
Which is a lovely thing to know, and also a slightly inconvenient one if you are trying to build a translator. There is no dictionary to look words up in, because there was never one book — there are millions of little ones, and each was written by one animal and one person over years of living together.
One sound we can be fairly sure about
Your cat has a hello.
The trill — that short rising chirrup, made with the mouth closed, half a second long and somewhere around 360 to 530 Hz — is the sound researchers read most consistently as a friendly greeting. Of everything in the cat repertoire it is the most reliably positive thing in there.
So if your cat trills at you when you come in, that is, as far as anyone can currently tell, a hello. Not a request. Not a complaint. A hello.
Dogs are louder and, oddly, harder. Pitch, harshness and above all the speed of repetition genuinely do carry information about how aroused a dog is and roughly what kind of situation it is in. But the literature contradicts itself about what a happy bark sounds like, and barks recorded from dogs left alone measure almost identically to barks recorded from dogs at play. So we hedge the dog more than the cat — and we tell you that we're doing it.
Which is why the app is careful
It would be easy to write software that puts sentences in an animal's mouth. Much of this category does exactly that: record a silence, receive a confident quote. Record the same silence again, receive a different confident quote.
We take the animal more seriously than that.
So PawParley tells you what it actually measured — how long the sound was, where the pitch started and where it ended, how far it moved on the way, how quickly it repeated. Then it offers a reading of what that vocalisation acoustically resembles, with its confidence showing. When the measurements point in different directions, it says so instead of picking one. And when there was no animal in the recording at all, it says that, and offers you nothing.
That restraint isn't a legal hedge. It's the part we're fondest of.
The whole strange pleasure of living with an animal is that it is a genuinely other creature, with an interior life you never quite get to read. An app that hands you a transcript is taking that away and giving you a magic trick in exchange. We would rather hand you the sound, honestly described, and leave the last step where it has always been — with the person who knows the animal.
We publish what we measured
Every result in the app shows its working, and we're happy to be held to it. We tested our own ideas the same way: we trained a mood classifier on published recordings, evaluated it properly with the individual cats held out of the test set, and it scored below the majority-class baseline. So it isn't in the app. What ships is the part we could stand behind.
There is no accuracy percentage anywhere in PawParley, and there won't be one. Nobody can truly translate an animal. That isn't a disappointment — it's the reason any of this is worth listening to.
The App Store link goes live at launch.