Silicon Valley pet emotional intelligence company Traini secures over 50 million yuan in financing to accelerate mass production of its first AI-powered smart collar

Text by | Fu Chong

Edited by | Su Jianxun

Silicon Valley-based pet emotional intelligence company Traini announced that it has completed a financing round of over 50 million RMB. The funds will be mainly used for the development of multimodal emotion models, the iteration of software and hardware products, and the expansion of overseas markets.

This round of financing was led by Banyan Tree, Silver Capital, ZhaoTai Group, and NYX Ventures, with follow - on investments from Strating Gate Fund, Jade Capital, and several technology investors, including two senior VPs from Nvidia, Julian Qian, a technical member of Anthropic, Zheng Weihe, the founding partner of Tongchuang Weiye, He Jia, the founding partner of Nanshan Capital, Peter Xu, the CEO of Plug and Play China, Zach Zhang, a partner at Edgewater Investments, etc.

The existing shareholder Tao Foundation and Hong Feng, the co - founder of Xiaomi, continued to participate in this round of financing.

Previously, Traini had received personal investments from executives of companies such as Google, Meta, and Palo Alto Networks. In May 2023, the company completed its angel - round financing. The investors included FutureX Capital, BlueSea Partners, Tao Foundation, Mint Capital, Valkyrie Fund, and Hong Feng, the co - founder of Xiaomi. They have been continuously investing in the exploration of algorithms and productization of pet emotional AI.

Traini's multimodal generative AI is applied to pet behavior understanding and "translation". Image provided by the company.

According to the company, Traini is one of the few teams currently applying multimodal generative AI to pet behavior understanding and "translation". The goal is to achieve near - real - time voice conversations between humans and pets based on Traini's proprietary multimodal model.

Currently, Traini's core software product is PEBI (Pet Empathic Behavior Interface). This system supports multimodal interactions across text, images, videos, and audio. By analyzing pets' barks, expressions, and behaviors, it estimates the psychological states of pets and outputs the results in a form similar to human conversations, enabling "conversations" between humans and pets.

Currently, Traini's model covers nearly 120 dog breeds, and the accuracy of emotion translation in internal tests can reach up to 94%. PEBI is open to pet clinics and smart hardware manufacturers in the form of an API, providing emotion and behavior understanding capabilities for medical and interactive devices.

This round of financing will accelerate the mass production of Traini's first AI smart collar. Image provided by the company.

After this round of financing, Traini plans to accelerate the mass production of its first smart wearable device - the Cognitive Smart Collar. This is the company's first generative AI - based hardware for human - pet interaction. Pre - orders are currently available on the Traini App and the official website traini.app.

This smart collar combines AI with real - time emotion tracking capabilities and has Traini's self - developed Valence–Arousal (VA) emotion model built - in. This model was trained by the Traini team in collaboration with pet behavior experts, based on more than 900 research papers on pet behavior and the behavior data of about 2 million dogs.

The collar collects physiological and behavioral signals such as heart rate, body temperature, and movement, and conducts comprehensive analysis in combination with barking patterns, converting the results into user - readable emotion and health profiles. The system can identify and track various emotional states and present the "translation results" to users in text or voice form, including the type of barks, emotional changes, and possible needs.

To support the above applications, Traini has also developed a 3D pet emotion model, using a custom immediate emotion vector to model the "internal emotions" of pets and the "external time and location context" in a unified way.

The company says that this model has three meanings: First, it makes pet behavior no longer an isolated action but a continuum of "emotion - habit - context"; second, it helps users identify the happy points, stress points, and potential risks of pets; third, it provides a basic framework of "behavior → emotion → context → suggestion → decision - making" for future pet smart wearables.

Traini's developed 3D pet emotion model uses a custom immediate emotion vector to model the "internal emotions" of pets and the "external time and location context" in a unified way. Image provided by the company.

On the user side, Traini has provided services (including tool use and model calls) for over 200 million pet dogs globally, and the playback volume of its related YouTube videos has exceeded 60 million. PetGPT, launched in 2023, supports natural language interaction and pet behavior analysis. Official data shows that the usage rate among covered users exceeds 99%, and the usage rate of related services has increased by about 70% through service matching.

In 2024, the company also released T - Agent, which can perform product searches, recommendations, and automatic orders based on Traini's model's understanding of dogs' needs. Currently, Traini has established partnerships with nearly 40,000 local pet stores in the United States.

Currently, Traini has built a high - quality pet audio - video dataset covering multiple breeds, scenarios, and emotional states, and is continuously optimizing the PPI system (Pet Perception & Interaction) based on fine - grained annotation. While maintaining recognition accuracy and generalization ability, the team adopts a lightweight and edge - deployment strategy to improve the response efficiency and privacy security of the product in a home environment.