Bark Mood Decoder

Read your dog's bark the way the research reads it

Not a translator. A wellbeing meter. It listens through your phone, reports a likely emotional state with a confidence score, and says "unclear" when the sound is ambiguous. It never turns a bark into a sentence.

Free to use No audio leaves your device No account Built in Layton, Utah
Ambient Signal reader
Arousal
-
Pitch
-
Texture
-
Closest published pattern
Start the microphone to read a live signal

This reader runs entirely in your browser. Nothing is recorded, uploaded or stored. It reports the three acoustic features the published bark research is built on, and names the pattern the sound leans toward. It is a signal reader, not the app's full model, and it is not a diagnosis.

Said plainly

What it does, and what it refuses to do

Most products in this category claim a translation they cannot demonstrate. We would rather tell you the shape of the thing and let you decide.

What it does

  • Listens to a bark or whine and reports a likely emotional state, such as playful, content, alert, anxious, demanding or assertive
  • Shows a confidence score with every read, and the second most likely state alongside the first
  • Says "unclear" and refuses to guess when the sound is ambiguous or the room is too noisy
  • Builds a log over time so you can see when your dog's normal changes
  • Runs on the phone you already own, with no collar and no hardware
  • Keeps every recording on your device, because the whole thing runs in your browser

What it does not do

  • Translate barks into words or sentences. Nothing in the peer reviewed literature supports that, so we do not sell it
  • Hold a two way conversation with your dog
  • Quote an accuracy percentage. No independent validation of this app exists yet, so we do not get to claim a number
  • Diagnose illness, pain or a behavioral disorder. That is a veterinarian's job and it always will be
  • Replace watching your dog. Body language carries more signal than sound does
  • Claim users, ratings or reviews. This is new and we have not earned those yet
How the reader works

Three acoustic features, from published research

Pongracz and colleagues found in 2006 that human listeners could classify dog barks by context, and that three acoustic properties carried most of that signal. The live reader above measures exactly those three and nothing else. You can watch it work.

Feature one

Pitch

The fundamental frequency of the bark. Low pitched barks skew toward aggressive and guarding contexts. High pitched barks skew toward fear, excitement and play.

The reader estimates pitch by autocorrelation on the raw waveform, in your browser.
Feature two

Texture

How tonal or how noisy the sound is. Harsh, noisy, broadband barks skew aggressive. Clean tonal barks and whines skew toward fear, anxiety and appeal.

Measured as spectral flatness across the frequency bins, live.
Feature three

Rhythm

The gap between barks. Short gaps and rapid repetition skew aggressive or alarmed. Longer, irregular gaps with higher pitch skew playful.

Counted from onset timing over a rolling window of a few seconds.
The honest caveat on this section

Those three rules are real findings about populations of dogs. They are not a verdict about your dog, in your kitchen, right now. A single sound can match a pattern and still mean something else. That is exactly why the app shows a confidence score and refuses to answer when the signal is weak.

The evidence

What the literature actually supports in 2026

Here is the current state of machine bark classification, with links, including the parts that are inconvenient for us.

Pongracz et al., 2006Journal of Comparative Psychology
Human listeners could classify dog barks recorded in different situations above chance, and specific acoustic properties mapped to specific contexts. This is the foundation the reader on this page is built on. PubMed record
University of Michigan, 2024LREC-COLING, Wav2Vec2 on barks
Reached up to 70 percent across four classification tasks on a dataset of 74 dogs, including playful versus aggressive. Individual dog identification came in near 50 percent. Not sentences, and not emotions in general. Paper and summary
Slobodian and Kozlenko, 2025Emotion classification preprint
Roughly 72 percent across five emotion classes, but from a single dog. The authors state directly that the accuracy "wasn't good enough for applications in the real world" and that it degraded across multiple dogs. Preprint
Sensors, December 2024Deep learning on dog vocalizations
Around 90 percent, but the task was distinguishing barking from howling in isolated environments. A much easier problem than reading emotion. Paper
The gap this leavesOur own reading of the above
The best peer reviewed canine vocalization work reaches roughly 70 to 72 percent on coarse categorical tasks. Nothing published supports a nineties-accuracy emotion translator across breeds. When you see that claim in an app store listing, ask who validated it.
And about our own numbers

Bark Mood Decoder has not been independently validated either. We are not exempt from the standard we just applied to everyone else. Until an independent validation exists, this page will not carry an accuracy percentage, and the app will keep showing you a confidence score instead of a promise.

Where this stops

The six things this product cannot do

Written down here so you never have to find out the expensive way.

Pricing

Free for owners, a working plan for professionals

The app is genuinely free. If you are a trainer, a shelter, a daycare or a vet practice and you need it shaped around your operation, that is the paid path.

$0

Bark Mood Decoder, the app

Open it in your browser, name your dog, and start reading. Unlimited reads, the mood deck, the log, the confidence scores, the whole thing. No account, no card, no download, and no audio leaving your device. There is no upsell inside it.

Open the free app
For trainers, shelters, daycares and clinics

Monthly working plans

These are Purpose Built AI monthly plans. You are buying build time and a direct line, applied to Bark Mood Decoder and to whatever else your operation needs.

Starter
$150 /month
For a solo trainer or a single location that wants the tool shaped to their intake.
  • A roster set up for the dogs you actually work with
  • A read log you can export and hand to a client
  • Small changes and fixes each month, on request
  • Email support with a same week response
Start Starter
Pro
$250 /month
For a practice running multiple dogs a day that needs the output to look like theirs.
  • Everything in Starter
  • Multi dog rosters and grouped logs
  • Your logo and wording on the client facing report
  • Priority on what gets built next
Start Pro
Scale
$400 /month
For multi location operations and shelters that need this wired into how they already work.
  • Everything in Pro
  • Multi location rosters and staff handoff
  • A report format built to your specification
  • Direct access to the person who builds it
Start Scale

Said plainly, because it matters: the professional dashboard is not a finished shipped product yet. What exists today is the working app you can open for free, plus a person in Utah who builds. A monthly plan buys that person's time and puts your operation first in the queue. If that is not what you want, use the free app and pay nothing. Cancel any month.

Questions

The things people ask first

No, and it never will. There is no published science supporting bark to sentence translation. What it does is report a likely emotional state and arousal level, with a confidence score, from the acoustics of the sound. If a product promises you sentences, that is a marketing decision, not a research finding.

We do not publish an accuracy number, because no independent validation of this app exists. Quoting a percentage we cannot back would be the single easiest way to lose your trust, and it is also the exact conduct the Federal Trade Commission acted on in its 2025 order against Workado over unsupported AI accuracy claims. Every read shows its own confidence instead.

Nowhere. The reader on this page and the app itself both run entirely in your browser. There is no upload, no server call, no account and no recording stored anywhere but on your own device. You can verify it by opening your browser's network tab while you use either one.

No. It uses the microphone in the phone or laptop you already have. The collar products in this category cost between roughly one hundred and several hundred dollars up front and then charge a monthly fee on top. This does not.

No. It can help you notice that your dog's vocal pattern has changed, which is a reasonable thing to raise with your veterinarian. It is not a diagnostic device and nothing it shows you should delay a vet visit.

Because the free app is built for one owner and one dog. A professional plan buys the version shaped around a roster, a log you can hand to a client with your name on it, and the ability to say what gets built next. You are paying for build time and access, not for unlocking features that already exist.

Purpose Built AI, a one person software studio in Layton, Utah, run by Jesse Beckstead. Small enough that you talk to the person who wrote the code.

Open free app Pro plans