The most unsettling AI products are not always the ones that look futuristic. Sometimes they look backward, toward someone a family is still trying to mourn.

In South Korea, AP reports that a growing number of people are turning to startups that create AI-generated videos of deceased loved ones. Seoul-based Vaice is one of the companies in that lane, using photos and voice samples to make short video messages that families often play at memorial gatherings or during major holidays.

This is not just another generative AI feature. It is grief becoming a consumer product.

The emotional pitch is easy to understand

The AP story opens with Lee Geon Hui, a 28-year-old office worker who commissioned an AI-animated video message from his late grandfather as a gift for his father. Lee wrote the message himself, trying to express feelings he believed his grandfather would have wanted to share. His father resisted watching it at first, then cried when he did.

That is the part of grief tech that cannot be dismissed with a cheap take. People miss parents, grandparents, partners, siblings, and friends. They want one more sentence, one more look, one more version of goodbye. AI companies have found a way to make that desire visible on screen.

Vaice CEO Jeongu Won told AP the company serves about 300 customers a month, mostly people in their 40s and 50s seeking videos of late parents. The basic three-to-five-minute video costs 600,000 won, or about $390, according to AP.

The technology is intimate by design

These products work because they borrow from the private archive: family photos, voice recordings, scripts, stories, and memories. That makes them different from a generic AI avatar or novelty filter. The whole point is emotional recognition.

AP reported that Vaice needs only a few photos and short voice samples to create a likeness. Other South Korean companies and TV programs have explored similar recreations of deceased entertainers, helping push the idea further into public view.

The appeal is obvious. So is the risk. Once a person's face and voice become usable after death, every family has to ask who gets to authorize that, who gets to write the words, and whether the dead should be made to say things they never actually said.

Consent gets messy after death

The hardest question is not whether the videos comfort people. Some clearly do. The harder question is who controls a person's likeness when they are no longer alive to say yes, no, or only under certain conditions.

A video created for a memorial ritual may feel loving inside one family. The same tool can feel invasive if relatives disagree, if the script rewrites a complicated relationship, or if a company later uses the dead person's image in ways the family never expected.

This is where AI grief videos start to overlap with the broader debate around deepfakes, digital rights, and posthumous identity. The law is still catching up to a technology that can turn a small archive into a moving performance.

Memory needs better infrastructure than hype

The practical lesson for families is not that everyone should make an AI version of someone they lost. It is that personal archives are becoming more powerful, and more vulnerable, than people realize. Photos, voice notes, home videos, letters, and scanned documents are now raw material for future tools, not just keepsakes.

For readers thinking about family history without jumping straight into synthetic resurrection, the safer lane is preservation: photo scanners, external backup drives, and archival photo storage. The goal should be keeping memories accessible, not rushing grief into the newest format.

The product category is only getting more intense

Right now, many AI grief videos are scripted clips. The next step is more interactive: chatbots, voice conversations, avatars that respond in real time, and systems that can imitate a person's style from larger archives of messages and media.

That is where comfort can slide into dependence. A short memorial video may help a family say goodbye. A permanently available simulation may make goodbye harder. The emotional design of these tools matters as much as the technical quality.

AI companies love to talk about productivity. Grief tech is a reminder that the real market may be broader and more fragile than that. The same systems that summarize meetings can also simulate absence. That is powerful. It is also a lot to put inside a checkout flow.

The future of mourning needs rules

AI grief videos are not going away. The desire behind them is too human, and the technology is moving too quickly. But the category needs boundaries: consent standards, family approval rules, clear deletion policies, limits on commercial reuse, and honest warnings about psychological risk.

The question is not whether people should be allowed to remember the dead with technology. People have always used technology to preserve memory. The question is whether AI companies can build products that respect grief instead of exploiting it.

That is the line Vaice and every grief-tech startup will have to walk. Memory can be a gift. It should not become just another subscription funnel.

As Amazon Associates, we earn from qualifying purchases.