Emotional Companionship

With fine-tuned large language models and highly customized speech models, you can easily transform the person, character, or entity you want into your virtual digital assistant. This allows you to have meaningful conversations with your favorite characters and receive assistance whenever you need it, greatly alleviating the loneliness of living abroad.

Just want to give it a try? Check out the demo video.

Digital Human Project

To create a highly immersive digital human experience, our product includes enhancements in the following three areas: language models (LLMs), text-to-speech models (TTS), and image generation models (DiM).

Important Notice

Please make sure to review the Privacy Policy.

Language Model

With your consent, we will use some of your conversation data to help train or fine-tune a personalized large language model. This will ensure that the assistant’s speaking style matches the character you envision.

Text-to-Speech Model

With less than one minute of audio data (though more is better), we can generate a voice model for your desired character. This makes the assistant's voice, tone, and intonation closely resemble the character you want. With two minutes or more of audio, the final result can be nearly indistinguishable from the original voice.

Here are two examples:

1. The character "Keqing" from the game Genshin Impact:

   Original Voice and Corresponding Text

   "I hope our show will bring joy to those who are looking forward to it."

   "I can't say for sure, but the possibility is indeed very high."

   Cloned Voice and Corresponding Text Example

   "With fine-tuned large language models and highly customized speech models, you can easily transform any character you want into your virtual digital assistant, which can greatly alleviate the loneliness of living abroad."

2. Developer "Yucheng":

   Original Voice and Corresponding Text

   "I am currently testing our study assistant's text-to-speech model."

   Cloned Voice and Corresponding Text Example

   "The assistant is built using a highly customizable multimodal large language model and rapid voice cloning text-to-speech technology, allowing your favorite character to accompany you as a virtual digital human during your studies abroad."

Image Cloning Model

Here, we only need a few photos of the character. The final generated image can have various styles, which you can refer to from the content in this link.

More Plans

You can provide feedback to us.