This AI pet collar claims it can translate your cat, dog, and others with 95% accuracy

A collar from Hangzhou is telling pet owners that the noise from the sofa, the hallway, or the food bowl may soon arrive as a sentence on their phones.

Pettichat, a product by Meng Xiaoyi, is a device that translates animal vocalizations and behavior to give an idea of what a pet might be feeling using short sentences.

It’s an audacious claim, but there is still little evidence. The AI system reportedly achieves around 95% accuracy, as per Meng Xiaoyi, and a figure of 94.6% has been reported by Chinese sources, but that’s for emotion detection rather than a direct animal dictionary translation.

Meng Xiaoyi pushes PettiChat into the pet AI market with a $118 preorder price

Pre-orders for the PettiChat launched in China by Meng Xiaoyi began on May 15, priced at 799 yuan ($118). Over 10,000 units have already been reserved, according to Chinese media.

PettiChat weighs approximately 27.2 grams and is therefore not an oversized collar but rather a standard pet accessory that will collect behavior information from pets. It includes built-in microphones that record sounds from animals. The system then analyzes behavior signals and generates a phrase for the user.

According to Dexerto, Meng Xiaoyi confirmed on the weekend that PettiChat leverages the Qwen model developed by Alibaba Cloud and records the voiceprints of pets. Alibaba Group’s stock is quoted under ticker BABA on the New York Stock Exchange and 9988.HK on the Hong Kong Stock Exchange. Alibaba claims that the solution will allow scientists to research the sound, behavioral, and emotional signals of voice.

It seems to be the right time for such a solution. Both pet care, wearables, and consumer AI sectors are booming. Additionally, Alibaba can leverage its huge domestic market. As domestic media note, the country will have 126 million pets among city residents by 2025.

The point here is that most pet owners are ready to pay for a deeper understanding of their four-legged companions. They observe their behavior carefully watching the bowl, tail, ears, pace, stare, and a very loud 2 am call that reminds them that it is time to pay rent.

Industry professionals warn against reading too much into application outputs

The challenging thing will be proving effectiveness outside a carefully controlled experimental setting. In the public record, there is no information on any independent studies or third-party testing done by Meng Xiaoyi for outsiders to examine. The lack of transparency in terms of data, testing methods, and the domestic environment raises serious concerns.

As the home environment is likely to be noisy anyway, the collar will pick up sounds of traffic, visitors, the television, other animals, music playing, vacuum cleaner operation, as well as the voice of an owner. While a system may function well in a laboratory, it cannot miss anything in a normal domestic situation.

Furthermore, animal behavior experts warn that animals do not communicate using their vocalization alone; they may convey various intentions depending on body posture, the position of the tail and ears, eye contact, walking pace, and other factors involved. For example, a bark when there is someone behind the door implies one set of intentions in a dog while a bark near an empty food bowl suggests another one.

This is precisely the area in which PettiChat may prove to be a nuisance. Should the application output generalized statements that coincide with owners’ own assessments, the buyer will think that they are paying for a pretty script. Otherwise, the application will continue working well enough for owners to overlook a lack of scientific data.

Meng Xiaoyi is also not selling PettiChat as only a sound toy. The pitch includes two-way communication, location tracking, and adaptive learning. That makes it look more like a wider pet-tech platform than a simple translation app.

The 799 yuan price tag makes things even more difficult. The consumer pays not only for the chance to locate the missing animal using the map function. They pay for the notion that the animal was always easy to understand, but only the smartphone could do it for them.