AI Kiss (Artificial Intelligence Kissing) is an interaction technology that employs multimodal interaction to simulate human kissing behavior through deep learning-based algorithms. It combines speech synthesis, haptic feedback, and image generation in order to simulate a virtual intimate experience. More than 12 million users have attempted to satisfy their emotional needs with AI Kiss technology worldwide, according to research conducted by the Human-Computer Interaction Laboratory at Stanford University in 2023. Of them, 35% are aged between 18 and 24, and on average, have a frequency of usage of 2.3 times a day and usage duration of 1.5 minutes per session on average. Technically, open-source libraries such as TensorFlow and PyTorch aid the developers with training customized models at a cost of less than $5. For example, using StyleGAN3 to produce virtual images of faces at 1024×1024 resolution, coupled with Meta’s Haptic Glove, pressure feedback accuracy is achieved at ±0.1N, and delay is controlled within 20ms.
Market data suggest that the free ai Kiss tool users’ growth rate was 47% in the second quarter of 2023, depending mainly on the monetization model through advertising. For example, Dreamlux AI video generator, a free model without watermarks, assists users in generating 1080p virtual kissing animations. Export time per run is only 45 seconds, the number of daily user generations is more than 30,000 times, and it doesn’t occupy additional GPU resources (video memory usage <2GB). Replika’s “Emotional Companion” function achieved an average month-long advertisement exposure of 15 times per individual by integrating cooperative brand content, and the website’s annual revenues have exceeded 80 million US dollars. With ElevenLabs’ s voice cloning API in the free plan, it offers a free quota of 10,000 characters every month to create a 10-second AI Kiss voice in only 0.3 seconds with 92% sound similarity. Canva’s library of AI design templates includes over 500 “virtual kissing” scene assets with a data usage of about 150MB per minute.
With regards to application scenarios, AI Kiss boasts a large penetration rate in healthcare, entertainment and social interaction. For instance, in 2022, the University of Tokyo employed VR+AI Kiss technology to heal patients with autism. The improvement rate in the social capabilities of the experimental group was 41%, and the cost of it was just 30% of that of traditional treatments. TikTok’s “AI Companion” filter contributed over 5 million downloads during its first week on the market. 23% of user-generated content (UGC) has the “aikiss” tag, and its average interaction rate is 18% higher than that of ordinary videos. In addition, the open-source community Hugging Face has also provided their AI model library with the pre-trained “BlowKiss-7B” model to enable developers to generate dynamic kissing animations for free using API calls, which have an inference time of merely 0.8 seconds.
Nevertheless, ethical controversies and technical limitations still exist. According to the European Union’s AI Ethics Committee 2023 report, untested AI Kiss tools have a 12% likelihood of generating inappropriate content, and there is also a 17% risk of user privacy leakage. For instance, in the initial testing of DeepMind, the haptic feedback error rate was as high as 8.5%, and out of the users, 7% suffered from motion sickness. But market research firm Gartner predicts that in 2025, the number of users of the world’s free AI Kiss tool will exceed 50 million, and the size of the market for emotional technology will be 7.4 billion US dollars. Of them, the share contributed by open-source tools such as Dreamlux AI video generator will exceed 60%. Its serverless architecture can reduce the cost per user to as low as $0.02 per invocation and make it capable of processing 150 requests per second, the most optimal for individual creators.