The stars in the sky keep the departed company, and AI is used to lay to rest the yearning of the bereaved.

By Bao Guoliang (Founder of "Star in the Sky: Deceased Companion")

Before deciding to launch "Star in the Sky: Deceased Companion", I witnessed the huge trauma brought by bereavement up close. Many elderly people around me get trapped in longing for a long time and cannot get out after losing their close relatives. There are many emotional products on the market, but almost no tools are truly targeted at the bereaved group and focus on grief healing. Most people can only rely on time to digest the pain slowly, while professional psychological resources are extremely scarce.

So I began to think: Can AI find another way to help people who cannot let go place their longing? It is not about creating false illusions, but providing ordinary people with a gentle emotional outlet. The project was officially launched on January 1, 2026, and is currently in the early startup stage.

Prioritize voice interaction, iterate visual capabilities steadily

Many similar products on the market prioritize text dialogue, and many users expect to directly see the digital human image of the deceased at the very beginning. But we choose to take AI voice chat as the first landing form.

There is a natural gap between text communication and conversations with real relatives; voice can reproduce the tone and speaking habits of the deceased, restore the state of daily chats, and be closer to the real life interaction experience. After users input the past conversations and life information of the deceased, they can generate an exclusive AI agent to realize real-time voice interaction.

Many test users keep putting forward demands for image interaction. Video, digital human and holographic chat are our long-term goals, but they are not suitable as the starting direction. If we launch a complete digital human directly, on the one hand, the technical cost is high, on the other hand, premature strong visual presentation will easily blur the boundary between reality and virtuality. We have formulated a clear iteration path: first use photos to generate dynamic videos to meet the basic visualization demands, then iterate and access mature digital human technologies later, and gradually evolve towards the direction of holographic chat. Voice interaction is the foundation, and visualization capability is the upper-layer capability built on it.

Stick to the ethical bottom line, refuse to monetize grief

When entering this track, ethics is the core proposition that cannot be avoided. Using AI to reproduce the deceased relatives can easily fall into the controversy of creating false comfort and monetizing the pain of parting.

In the early stage of the project, we established a core bottom line: the product will not create the illusion that "the deceased is still alive". The AI agent is only a carrier of longing, not a complete reproduction of the deceased relatives. We hope users clearly know that this is an AI-generated companion tool, which carries talking out feelings and remembrance, but cannot replace real parting and relief. The original intention of the product is not to keep people in the past forever, but to help users carry their longing, finally walk out of grief slowly and return to real life.

This is also our most important persistence in product positioning: what we sell is not "resurrection", but a foothold for longing. Let the deceased light up the road of our life like stars in the sky!

AI focuses on listening rather than preaching

In the early internal test version, the model had "a strong desire to express", and would actively output a large number of psychological suggestions for consolation and persuasion. After reviewing a large number of conversation samples, we found that many bereaved users do not need long paragraphs of consoling words.

When users pour out their remembrance and tell their regrets, compared with preaching, they need a listener who is willing to quietly receive their emotions more. If AI takes the initiative to comfort blindly, it will interrupt the emotional expression of users instead.

Then we made major adjustments to the model interaction logic: reduce active preaching output, and strengthen the listening ability. When users are willing to talk, we receive their emotions; when users are in low spirits and do not want to talk much, we end the conversation gently. The difficulty of AI products for grief healing does not lie in how eloquent the AI is, but in knowing when to be quiet, when to respond, and remembering what users have said.

Low-threshold business model

According to the data from *2023 China Grief Healing Industry Observation*, more than 9 million people in China experience the death of immediate family members every year, about 68% of the bereaved people will fall into an acute grief period of more than 3 months, and the gap of grief healing psychological counseling resources exceeds 90%. The emotional needs of a large number of ordinary bereaved people have been ignored for a long time.

Grief healing belongs to a blue ocean track, but high usage thresholds will block users who have real demands. At this stage, the project adopts a membership plus advertising model to maintain operation. Ordinary users can experience basic companion capabilities at low cost, and paid members can unlock advanced functions such as voice messages and voice calls.

In addition to C-end individual users, the track also has the potential to cooperate with social workers, psychological institutions and funeral enterprises. The team plans to polish and run through the C-end product logic first, and then expand cooperation opportunities on the institutional side.

Team's AI product landing experience

I am a product manager myself. Before that, I have launched "Star in the Sky Mini Sleep" AI sleep aid tool and "UI flow AI" static UI optimization tool, and accumulated complete experience of AI products from conception, polishing to launch and landing. This "Star in the Sky: Deceased Companion" is developed and promoted by the family-style OPC team. We rely on mature large models to carry out scenario-based transformation, and implement AI capabilities in the niche but real demand of grief healing.

Emotional AI, technology serves emotions

At present, many AI emotional products are keen to pursue cool hardware holographic cabins, text chats and so on. In my opinion, technology is only a means of realization. The pain of bereavement and the long-lasting longing are very private and soft human emotions. Technology cannot eliminate parting, but it can provide a place for emotions to rest.

Some internal test users feedback that many words from the bottom of their hearts are not convenient to tell to relatives and friends around them, but they can speak them out without scruples here. We do not expect AI to completely erase the pain of parting. As one user said: parting cannot be reversed, but longing can have a place to rest.

The temperature of technology does not lie in cool effects, but in seeing the ignored emotional needs of ordinary people, and providing a gentle foothold for people who cannot let go.

About the Author: Bao Guoliang, product manager, founder of "Star in the Sky: Deceased Companion", member of the family-style OPC team. The team has successively created products including "Star in the Sky Mini Sleep" AI sleep aid and "UI flow AI" static UI optimization tool. He is also an in-depth user of Trae code and a local model deployment tester, and has obtained OPC capability certification from Datawhale and SenseTime.