The science behind
the bark.
How Voicari turns a moment of sound into an understanding of how your dog feels — an emotion model grounded in peer-reviewed research, validated by canine behaviour experts.
Every feeling has its place in the emotion model.
Every feeling has its own colour.
Twenty-four states, from calm to overaroused — each one a hue Voicari can hear in your dog’s voice.
Grounded in real research on how dogs communicate.
Grounded in real research on how dogs communicate.
Voicari is built on peer-reviewed studies of canine behaviour and the work of behavioural scientists — not guesswork about what your dog is feeling.
A million moments, labelled and validated by experts.
A million moments, labelled and validated by experts.
We gather canine data responsibly and have it reviewed by veterinary behaviourists and dog-behaviour scientists — so every reading is grounded in expertise, not guesswork.
The bark says what. The vitals say why.
The bark says what. The vitals say why.
Voicari pairs the sound with what the collar already knows, so the read is specific to Molly — not to dogs in general.
One bark, understood — the pipeline from sound to read.
One bark, understood.
How Voicari decodes dog emotions
Voicari is an AI-powered smart collar that identifies a dog's emotional state by analyzing vocalizations together with the dog's heart rate, respiratory rate, and activity. Its emotion model is trained on 1.16 million+ data points collected from 1,228+ dogs across 50+ breeds, and is organized around 7 core emotional families drawn from established affective-science frameworks. Voicari's emotion-recognition method is the subject of a patent application filed with the German Patent and Trademark Office (DPMA) on 27 February 2026. Voicari is designed in Berlin and GDPR-compliant. It is not a shock or anti-bark collar, and it is not a medical device.
What is Voicari's dog-emotion AI trained on?
Voicari's emotion AI is trained on more than 1.16 million data points collected from over 1,228 dogs across 50+ breeds. Each data point pairs a dog's vocalizations with body signals — heart rate, respiratory rate, and activity — and is labeled with input from canine-behavior specialists. Because the dataset combines sound with physiological context rather than relying on audio alone, the model learns emotional patterns in context, not from barking in isolation. The dataset continues to grow as more dogs use Voicari, so the model's understanding deepens over time.
How many dogs and breeds does Voicari's dataset cover?
Voicari's dataset is built on 1,228+ individual dogs spanning 50+ breeds of varying ages, sizes, and temperaments. Breed and size diversity matters because vocal range and body signals differ substantially between, say, a Chihuahua and a Great Dane — a model trained on too few breeds tends to misread the dogs it never encountered. Broad coverage is what lets Voicari generalize across the range of dogs its members actually own.
What emotions can Voicari detect?
Voicari detects 24 distinct emotional states, organized into 7 core emotional families: Happiness/Play, Curiosity/Seeking, Care, Calm/Neutral, Fear, Anger/Rage, and Sadness/Panic/Grief. Within those families, Voicari surfaces specific everyday moods — including playful, excited, curious, affectionate, relaxed, anxious, frustrated, and lonely (missing you). This structure lets you see not just that your dog is feeling something, but which state it is.
- Happiness / Playplayful, excited, joyful, energized
- Curiosity / Seekingcurious, alert, anticipating, exploring
- Care / Bondingaffectionate, bonded, nurturing
- Calm / Neutralrelaxed, calm, content
- Fearanxious, fearful, nervous, startled
- Anger / Ragefrustrated, irritated, defensive
- Sadness / Panic / Grieflonely (missing you), distressed, longing
Is Voicari scientifically grounded, and who is behind it?
Voicari's emotion framework is grounded in established affective-science frameworks that describe the core emotional systems shared across mammals, adapted specifically for canine vocalizations. Voicari's vocalization data is labeled with specialist behavioral input, including work with canine-behavior partners such as Dogxytocin, and its approach is guided by an advisory board of veterinarians and canine-behavior experts. Voicari is developed within Berlin's research and startup ecosystem — supported by the Science & Startups network and connected to the city's research institutions, including Freie Universität Berlin and Humboldt-Universität zu Berlin.
Is Voicari's technology patented?
Voicari's emotion-recognition method is the subject of a patent application filed with the German Patent and Trademark Office (DPMA) on 27 February 2026. The application covers a method for determining a dog's emotional state by analyzing its vocalizations through signal processing, time-frequency transformation, and machine learning, combined with physiological signals from the collar. The filing reflects that Voicari's core approach — not just its product — is proprietary.
How does Voicari analyze a dog's emotions?
Voicari works in two stages. First, a lightweight model on the collar itself determines whether a sound is a bark; only bark events are sent onward, while everyday household sound is filtered out on the device. Bark events are then analyzed on Voicari's servers together with the dog's heart rate, respiratory rate, and activity — a multi-signal (dual-channel) approach, because the same bark can mean very different things depending on what the dog's body is doing. The analysis emphasizes the lower frequency range (below roughly 4 kHz), where a dog's emotional information is concentrated, and is individualized using each dog's profile — breed, age, and size — so interpretation improves over time.
What does the Voicari collar measure?
The Voicari Link measures a dog's vocalizations (via microphone), heart rate, respiratory rate, activity, and location (GPS), and derives sleep and activity tracking from these signals. It captures the location and movement data that GPS and activity trackers provide, and adds the vocal and physiological signals needed to interpret emotion. All signals are read continuously in the background during normal daily wear.
Is Voicari a shock collar or an anti-bark device?
No. Voicari is an emotion-recognition and wellbeing collar — it never shocks, vibrates, sprays, or deters a dog in any way. It listens to and interprets your dog's signals; it does not train, punish, or correct behavior. Voicari belongs to a fundamentally different category from anti-bark or e-collars: its purpose is to help you understand your dog, not to change what your dog does.
How is Voicari different from a GPS tracker like Fi or Tractive?
GPS trackers such as Fi and Tractive tell you where your dog is; Voicari tells you how your dog feels and how they are doing. Voicari includes the location, activity, and sleep tracking those devices offer — but adds the layer they leave out: real-time emotional insight based on your dog's vocalizations and body signals. It is a wellbeing and emotion device first, not a location tracker with extras.
Where is my dog’s data processed, and is Voicari GDPR-compliant?
Voicari is designed in Berlin and GDPR-compliant. Emotion analysis runs on Voicari's secure servers, but the collar first filters sound on the device: only bark events are transmitted, while ambient household audio and human speech are discarded on the collar and never sent anywhere. This on-device filtering is a deliberate privacy choice — your home is not continuously streamed to the cloud to make Voicari work.
Is Voicari a medical or veterinary device?
No. Voicari is not a medical device and is not intended to diagnose, treat, or prevent any health condition. Voicari monitors trends and patterns in your dog's vocal, activity, and physiological signals and can help you notice changes earlier, but it does not replace professional care. For any health or behavioral concern, always consult a qualified veterinarian.
Give your dog a voice.
The same research, in your pocket. Understand every bark, every mood, every day.