How accurate is FaunaVox?
Every number on this page was measured on recordings from people FaunaVox never learned from, including the animals it gets wrong. FaunaVox 1.2.0.
62.5%
of animal sounds named right, with the free extra-accurate pack (24.4% without it)
73.6%
of whole recordings named right, with the pack (26.5% without)
88.2%
of “likely” answers were right (84.8% without the pack); “possible” about 49.6%
9.2%
of everyday non-animal sounds were mistaken for an animal (10.4% without the pack)
How we measure
- 6,976 recordings, 30,433 sound events of 114 animals and everyday non-animal sounds, from Wikimedia Commons, iNaturalist, Freesound, the U.S. National Park Service and NOAA (all openly licensed).
- Held out by recording session. Every recordist (an iNaturalist user, a Freesound user, a Commons uploader) is one session; a sound is only scored by a model that never heard anything from that session. Re-uploaded copies of the same recording were found by comparing the audio and removed.
- The app's own code. Results are scored with the same interpreter the app runs, in Wild Mode, region “Anywhere”.
- Confidence is calibrated on these held-out sounds, and capped at 90%: an animal-sound app should never claim to be certain.
Pets: the kind of call
Whether a bark is a bark or a whine decides the mood FaunaVox shows. Right call type on recordings from people the call models never trained on:
| Animal | Standard | With the pack |
|---|---|---|
| Dogs: bark, whine, howl, growl | 80% | 83% |
| Cats: meow, purr, hiss, yowl | 79% | 83% |
| Chickens | 70% | 70% |
A dog's play growl and warning growl could not be told apart by sound, so FaunaVox shows one “growl” and asks you to look at the body language. A warning growl is only ever changed when the model is at least 80% sure.
What FaunaVox cannot do (yet)
- Meanings are likely interpretations from animal-behaviour research, never word-for-word translations. Experts agree on what a call means only about 70–80% of the time. There is no ground truth for what a dog “actually meant”, so any app claiming 95% translation accuracy is measuring something else.
- The mood map (pleasant ↔ unpleasant, calm ↔ excited) is estimated from the kind of call, its pitch and how fast it repeats. It follows published research on animal emotion, but it has not been tested against independent ratings yet.
- Health sounds: on 41 recordings of pets, FaunaVox caught 10 coughs, sneezes, snores or pants, with 1 false alarm in 140 animal recordings. So it only ever says “possible” and asks you, and it never diagnoses.
- Smoke alarms and breaking glass: caught in 15 of 38 alarm recordings and 25 of 50 glass breaks. An extra warning, never a replacement for a smoke detector.
- Danger alerts: 177 of 439 sounds of dangerous animals raised a warning; harmless sounds raised a full-screen alarm 0.1% of the time. Never rely on an app alone for safety.
- The “second opinion” from AI on your device that listens to the clip is not measured by us. The app's test bench measures it on your own device and shows the result beside every second opinion.
- Bears, elephants, cougars and hyenas are still rarely recognised, and whales and dolphins need an underwater microphone.
Every animal
Share of that animal's test sounds named right (exact animal), without and with the extra-accurate pack.
| Animal | Test sounds | Standard | With the pack |
|---|---|---|---|
| Alligator or crocodile | 36 | 2.8% | 16.7% |
| Asian koel | 291 | 32.3% | 70.8% |
| Australian magpie | 208 | 2.4% | 59.6% |
| Babbler | 98 | 2% | 34.7% |
| Barbet | 278 | 4.7% | 62.2% |
| Barking deer (muntjac) | 198 | 26.8% | 65.7% |
| Barn swallow | 273 | 13.6% | 84.2% |
| Bear | 63 | 3.2% | 3.2% |
| Bee-eater | 221 | 2.3% | 52.5% |
| Bees or wasps | 200 | 70.5% | 75.5% |
| Bellbird (korimako) | 269 | 3.3% | 63.6% |
| Black kite | 163 | 0% | 43.6% |
| Blackbird or thrush | 483 | 17.4% | 73.7% |
| Blue jay | 216 | 5.1% | 71.3% |
| Bulbul | 367 | 5.2% | 58.3% |
| Canary | 191 | 6.3% | 64.9% |
| Cat | 1397 | 73.8% | 76.3% |
| Chicken | 415 | 72.8% | 75.2% |
| Cicada | 360 | 29.2% | 76.7% |
| Common cuckoo | 274 | 18.2% | 82.5% |
| Common loon | 201 | 18.4% | 69.2% |
| Common pheasant | 186 | 4.8% | 59.7% |
| Common raven | 232 | 5.2% | 64.7% |
| Coucal (crow-pheasant) | 120 | 5.8% | 49.2% |
| Cougar (mountain lion) | 11 | 0% | 0% |
| Cow | 289 | 66.8% | 69.2% |
| Coyote | 399 | 13% | 59.4% |
| Crane | 111 | 5.4% | 79.3% |
| Cricket | 415 | 64.6% | 85.5% |
| Crow | 310 | 64.2% | 71% |
| Dog | 1212 | 75.1% | 72.3% |
| Dolphin | 76 | 7.9% | 35.5% |
| Donkey | 66 | 30.3% | 74.2% |
| Drongo | 223 | 0% | 34.1% |
| Duck | 358 | 57.3% | 64.2% |
| Eagle | 226 | 2.7% | 48.2% |
| Eastern whipbird | 231 | 12.6% | 60.2% |
| Elephant | 87 | 1.1% | 3.4% |
| Elk | 188 | 8% | 39.4% |
| Frog or toad | 680 | 59.1% | 72.9% |
| Gecko | 128 | 0% | 53.9% |
| Goat | 113 | 56.6% | 53.1% |
| Goose | 582 | 58.8% | 75.9% |
| Grackle | 231 | 2.2% | 50.6% |
| Grasshopper | 186 | 15.1% | 68.8% |
| Gray catbird | 256 | 0% | 76.2% |
| Gray wolf | 124 | 14.5% | 31.5% |
| Great kiskadee | 193 | 4.7% | 71% |
| Guinea fowl | 88 | 21.6% | 70.5% |
| Guinea pig | 118 | 33.9% | 74.6% |
| Gull | 301 | 9% | 70.4% |
| Hawk-cuckoo (brainfever bird) | 104 | 21.2% | 62.5% |
| Hedgehog | 83 | 8.4% | 12% |
| Heron | 210 | 8.1% | 65.7% |
| Hoopoe | 236 | 20.3% | 78.8% |
| Horse | 143 | 56.6% | 62.2% |
| House sparrow | 371 | 25.6% | 87.3% |
| Humpback whale | 202 | 27.7% | 45.5% |
| Hyena | 51 | 0% | 3.9% |
| Indian peafowl | 100 | 7% | 58% |
| Indigo bunting | 207 | 4.3% | 65.2% |
| Jackal | 108 | 4.6% | 41.7% |
| Killdeer | 230 | 11.7% | 72.2% |
| Kingfisher | 186 | 3.8% | 43% |
| Lapwing | 257 | 5.4% | 58.4% |
| Laughing kookaburra | 286 | 19.6% | 67.5% |
| Magpie | 343 | 7% | 65.9% |
| Meadowlark | 251 | 5.2% | 61% |
| Mockingbird | 336 | 3.9% | 61.6% |
| Monkeys and apes | 695 | 15.5% | 46.2% |
| Moorhen, coot or waterhen | 254 | 2.8% | 53.9% |
| Myna | 238 | 0% | 39.1% |
| Nightjar or whip-poor-will | 244 | 7.8% | 76.6% |
| Noisy miner | 296 | 0.3% | 54.4% |
| Northern cardinal | 238 | 3.8% | 77.7% |
| Nuthatch | 277 | 0% | 47.7% |
| Orca (killer whale) | 31 | 6.5% | 25.8% |
| Oriole | 250 | 0.4% | 45.2% |
| Owl | 416 | 19% | 56% |
| Parrot or budgie | 755 | 16.7% | 58.3% |
| Phoebe or pewee | 209 | 0.5% | 57.4% |
| Pied currawong | 241 | 3.3% | 60.6% |
| Pig | 156 | 53.8% | 56.4% |
| Pigeon or dove | 295 | 36.3% | 59.3% |
| Prairie dog | 59 | 6.8% | 71.2% |
| Prinia | 224 | 0.4% | 54.5% |
| Quail | 185 | 0.5% | 54.1% |
| Rabbit | 37 | 29.7% | 40.5% |
| Rainbow lorikeet | 272 | 5.9% | 66.2% |
| Rattlesnake | 86 | 10.5% | 48.8% |
| Red deer | 120 | 3.3% | 21.7% |
| Red fox | 290 | 10% | 45.2% |
| Red-tailed hawk | 187 | 1.1% | 58.3% |
| Red-winged blackbird | 256 | 1.6% | 60.5% |
| Rose-ringed parakeet | 290 | 0.7% | 63.1% |
| Sambar deer | 11 | 0% | 0% |
| Seal or sea lion | 103 | 11.7% | 29.1% |
| Sheep | 242 | 69.8% | 78.9% |
| Snake | 20 | 25% | 15% |
| Songbird | 357 | 71.7% | 82.6% |
| Sperm whale | 43 | 0% | 2.3% |
| Squirrel | 314 | 11.1% | 36.9% |
| Stork | 54 | 20.4% | 55.6% |
| Superb lyrebird | 361 | 15.8% | 69.8% |
| Swan | 184 | 6% | 52.2% |
| Tanager | 299 | 23.7% | 76.6% |
| Tiger or lion | 156 | 64.1% | 67.3% |
| Toucan | 224 | 24.6% | 87.1% |
| Towhee | 251 | 2% | 64.9% |
| Treepie | 190 | 2.1% | 46.3% |
| Tūī | 324 | 4.3% | 50.9% |
| Turkey | 267 | 34.1% | 50.6% |
| Vireo | 310 | 5.8% | 69.4% |
| Woodpecker | 572 | 1.2% | 40.6% |
Test it yourself
Open FaunaVox → Learn → Test it yourself. Name your recordings after the animal (for example domestic_dog__whine__01.wav), choose them, and the test bench scores them exactly like the numbers above, on your phone, without uploading anything. It exports the results as a spreadsheet.
Researchers and vets who would like to test FaunaVox independently are very welcome: write to us through the contact address on the Google Play listing and we will share the test protocol.