Before speech, before thought, before anything that makes us human — there is breath. A dog does not hide its breathing. It cannot. Every inhale and exhale is an honest signal of vitality, distress, or disease.
A veterinarian places a stethoscope on a dog's chest. They hear harsh breathing. They note it. But they cannot answer the question that matters: Is this harsher than it should be for a Pug breathing 28 times per minute after a walk?
No hardware. No wearable. No subscription. Just a phone and a dog.
You hold a phone near your dog, press record for 8 seconds, and the app separates the dog's breathing from the room noise, finds each sound event, measures it, and compares it against what is expected for a dog of this breed, age, and breathing rate.
Nine layers. Every step inspectable. Nothing hidden.
8 seconds at 48,000 samples per second. 1.5 seconds of pre-roll captures the room's noise fingerprint. Four audio flags flipped to preserve signal integrity.
Spectral subtraction removes the room's noise signature. Adaptive filtering separates the dog's breathing from HVAC hum, traffic, conversation.
Signal-to-noise ratio, clipping, frequency rolloff, gain control suspicion. Below threshold: metrics are withheld, not computed on garbage. The instrument says "I won't measure this" rather than lie.
Energy envelope tracked in 10-millisecond frames. Adaptive threshold finds where sound events begin and end. Events merged if close, discarded if too brief.
Two-stage architecture. First: statistical pattern recognition against thousands of reference recordings. Second: deterministic physics validator checks whether the prediction matches expected spectral signatures. When they agree: high confidence. When they disagree: the instrument admits uncertainty.
Eleven measurements computed where applicable: pitch, harmonic-to-noise ratio, spectral centroid, formants, jitter, shimmer, respiratory rate, inspiration-to-expiration ratio, and more. Wrong event type returns null — no false precision.
Harshness Index: measured harshness minus expected harshness for this breed, age, weight, and respiratory rate. Drift: current recording compared to the pet's own baseline on the same device — the phone's distortion cancels out mathematically. Reference Correlation: percentile ranking against a global database of validated recordings.
Deterministic rules produce structured clinical observations. Not "your dog has pneumonia." Rather: "Your recording sits at the 87th percentile for cough harshness among resting dogs, consistent with patterns seen in pneumonia and kennel cough profiles."
Fourteen synthetic tests with known inputs assert measured precision. External validation against seventeen reference datasets. The instrument spec sheet is published, not claimed.
Fauna Acoustic is not a dataset company. It is a standardisation and correlation layer.
Google didn't build the web. They indexed it. Bloomberg didn't build markets. They standardised the data feed. Fauna Acoustic doesn't build the acoustic dataset. It standardises the analysis pipeline that makes all existing data clinically useful.
| Competitor | What They Do | Our Differentiation |
|---|---|---|
| BrachySound | BOAS detection in Pugs. 0.68 accuracy. | They do ONE condition in ONE breed. We do 9 event types across all breeds with breed scaling. |
| AudioSet / Barkopedia | Raw audio datasets with labels. | They have data. We have the clinical pipeline that makes it useful. |
| PetPace / Whistle | Wearable health monitors (activity, temp, HR). | They measure physiology. We measure acoustics — the only modality that captures respiratory events. |
| Digital Stethoscopes | Hardware devices for vets. | We are software-only. Works on any phone. No hardware purchase. |
The pipeline runs end-to-end on an iPhone 15 Pro. First real-world recording analysed: Leo, a diabetic Pug, at rest.
Primary use case: A clinician opens Fauna Acoustic to assess severity of pneumonia, collapsed trachea, or irregular breathing patterns not evident in preliminary checks. A pet parent records after noticing something odd — cough, unusual panting, lethargy. Per-usage analysis, not daily monitoring.
Why this fits: Daily monitoring requires habit formation (hard). Per-usage analysis requires only one moment of concern (easy). The output is "data for your vet," not "your dog has pneumonia." This is the difference between a Fitbit and an ECG machine.
V1 demo deployed. Working web app. Record → analyse → structured readout. Clinical validation begins with veterinary partners.
One veterinary clinic validates outputs against their own assessments. The corrections schema seeds the learning corpus for V2. Reference database expands to 1,000+ fingerprints.
Telemedicine platforms integrate Fauna Acoustic as an API. Pet insurance companies use it for risk assessment. Veterinary clinics use it as a dashboard instrument.
Integration with wearable devices. Continuous monitoring without the phone. The same acoustic analysis, now autonomous.
A purpose-built veterinary instrument with calibrated microphone, active noise cancellation, and real-time analysis.
Bodh designs and manufactures its own wearable device. End-to-end control of hardware, software, and data pipeline.
Do you know a veterinary researcher, a clinic owner, or an IP attorney who works in biotech? Warm introductions matter more than cold emails.
Where does this pitch fall apart? Which assumption is weakest? We need honest pushback, not validation.
You have seen products succeed and fail. Does this feel like a real problem with a real solution, or a solution looking for a problem?
If this resonates, we are looking for advisors and early partners. Not capital yet — validation and network.
A phone becomes a calibrated stethoscope. It measures what the human ear cannot count — and what veterinary medicine abandoned because no tool existed.
Record your dog. See what the instrument hears.
Try NowOpens in new tab · fauna-acoustic.vercel.app