CONSUMER

Seance AI

A GPT-4 Grief Tech App That Lets You Speak with Lost Loved Ones

AE Studio built Seance AI, a grief tech application that uses large language models to simulate text-based conversations with deceased loved ones. Designed for emotional closure rather than ongoing use, the app attracted national press coverage from Fox News, Futurism, and The Sun at launch.

Seance AI

THE CHALLENGE

The problem.

Grief is one of the most universal human experiences, and for centuries people have sought ways to find closure after losing a loved one. Traditional counseling and journaling each serve a purpose, but none address the specific longing for one final conversation with someone who is gone.

With the rapid rise of large language models in 2023, a new possibility emerged: could AI simulate a person's communication style well enough to offer a meaningful, if brief, sense of closure? And if so, how should that experience be designed responsibly?

Most grief tech companies in this space positioned their products as preservation tools, carefully avoiding any language that implied resurrection. AE Studio took a different approach. Rather than obscuring what the technology was doing, the team leaned into the idea directly, naming the product Seance AI and treating transparency as a core design principle.

The challenge was building something emotionally resonant without being exploitative, technically credible without overclaiming, and honest enough about its limitations that users could engage with it in a healthy way.

THE SOLUTION

What we built.

Radical Transparency as a Design Principle

Most grief tech products soften the premise. Seance AI did the opposite. The name was chosen deliberately to signal exactly what the product was: a fictionalized, AI-generated conversation with someone who is no longer alive.

AE Studio explained that the name was intentionally striking. The goal was to draw attention to how advanced large language models had become and to give users a clear-eyed frame for what they were engaging with. A product that admits it is conjuring a digital simulacrum is more honest than one that implies it has captured a soul.

Personality Modeling Through User Input

Seance AI collected a structured set of inputs to build a personality model for the deceased. Users provided the person's name, age, cause of death, personality traits rated on a 1-5 scale, examples of the person's writing style, their relationship to the user, and subjects they wanted to discuss.

Writing samples were the most important input: an old email, a letter, or a text message gave the model a template for vocabulary, tone, and emotional register. The result was a chatbot that could approximate how someone communicated, not just what they might generically say.

GPT-4 Integration and Conversational Interface

The application was powered by GPT-4 via OpenAI's API. Once onboarding inputs were submitted, users were brought to a minimal conversational interface preceded by an animated flame visual that signaled the transition into the seance.

The experience was designed to feel appropriately atmospheric without being overwrought. From there the interaction was straightforward: a text box, a flame, and space for the conversation to unfold.

Intentional Brevity and Closure-Focused Design

Rocks was explicit that Seance AI was not meant for long-term or repeated use. He compared it to a traditional seance: a bounded ritual, not an ongoing relationship. The product was designed to facilitate closure and emotional processing in a single session, not to create a persistent digital version of the deceased.

To reinforce this, the team moved away from a monthly subscription model toward a pay-per-session pricing structure. The logic was direct: making it easy to summon the dead repeatedly would undermine the product's core purpose.

Premium Features: Voice Recreation and Animated Images

Beyond the base text experience, Seance AI's premium tier introduced voice recreation and animated images. For voice, users could upload up to three audio samples of the deceased, each under five minutes, which the system used to reconstruct a voice model.

The animated image feature added subtle movement to photos of the deceased, including gentle eye and head motion. Together these features pushed the experience toward something richer and more sensory while remaining grounded in what the technology could responsibly deliver.

AI Safety as Part of the Product Narrative

Rocks and the AE Studio team were deliberate about grounding the product in a public conversation about AI safety and ethics. Rather than avoid the uncomfortable questions the product raised, they used its visibility to advocate for thoughtful regulation and awareness of what LLMs can and cannot do.

Rocks told Fox News that AE Studio was greatly concerned about AI safety and wanted to draw attention to the potential implications of the technology, while also making clear they supported healthy regulation rather than halting progress. Seance AI was designed to be a demonstration of where AI was heading and an invitation to engage seriously with what that means.

HOW IT WORKS

The details.

Leading With Transparency Instead of Euphemism

The product was intentionally named Seance AI to make the experience explicit from the first interaction. Instead of framing the system as preservation software, the team presented it as a clearly simulated conversation. That transparency shaped product copy, onboarding expectations, and user trust from the start.

Building a Personality Model From Structured Inputs

Users provide key context about the deceased person, including personality traits, writing samples, and relationship details. The writing samples act as the strongest signal for tone and vocabulary. This input package gives the model enough context to generate responses that feel specific to one individual rather than generic chatbot output.

Running the Conversation on GPT-4 With a Focused Interface

After onboarding, the app moves users into a minimal conversational experience powered by GPT-4 through OpenAI's API. The interface stays simple by design: atmospheric transition, text input, and message flow. The implementation favors clarity and emotional readability over feature-heavy interaction patterns.

Designing for Closure With Explicit Usage Limits

The product was designed as a short, bounded experience rather than an always-on relationship. Pricing shifted toward pay-per-session to reinforce one-time or occasional use. Premium capabilities, including voice recreation and animated images, were added as optional enhancements without changing the core closure-first interaction model.

OUTCOMES

What shipped.

Featured in Fox News, Futurism, and The Sun at launch

Futurism reviewer found the first AI response approximately 80% convincing in replicating her late father's communication style

Coverage reached technology, mental health, and spirituality-adjacent audiences beyond typical tech press

Sparked public discourse on AI ethics and grief technology cited across multiple publications

KEY TAKEAWAYS

What we learned.

  • Radical transparency outperforms euphemism. Naming the product Seance AI and leaning into what it actually did created more trust and more press interest than a softer framing would have.
  • Closure-oriented design requires explicit constraints. Building in session brevity and pay-per-session pricing were as important as the technical implementation. The design philosophy was inseparable from the ethics.
  • Writing samples are the most important input for personality simulation. User-supplied text from emails, letters, or messages gave the model the vocabulary, tone, and emotional register needed to approximate a specific individual.
  • AI safety should be part of the product narrative from day one. Addressing limitations openly, rather than defensively, built credibility and positioned the team well in a rapidly evolving public conversation about LLMs.
  • Skunkworks projects can generate disproportionate visibility. A tightly scoped internal build attracted national press and positioned AE Studio as a thought leader in a category that barely existed at the time.

IN SUMMARY

Bottom line.

Seance AI was AE Studio's exploration of one of the most emotionally charged applications of large language models: simulating a conversation with someone who has died. By leading with transparency, designing for closure rather than dependency, and engaging seriously with the ethical dimensions of the product, the team built something that resonated far beyond its technical scope.

The national press coverage and the ongoing conversation the product sparked pointed to the same conclusion: building at the frontier of AI, done thoughtfully, creates real value. Seance AI remains a demonstration of what is possible when technical capability and ethical clarity are treated as complementary rather than competing forces.

FAQ

Frequently asked.

What is Seance AI and who built it?
Seance AI is a grief tech application built internally by AE Studio. It was created by AE Studio, and launched in May 2023. The app uses GPT-4 via OpenAI's API to simulate text-based conversations with deceased loved ones, based on personality information and writing samples provided by the user.
How does Seance AI simulate a deceased person's communication style?
Users provide a structured set of inputs including the deceased's name, age, cause of death, personality traits rated on a 1-5 scale, their relationship to the user, and most importantly, examples of how the person wrote. These writing samples, pulled from old emails, letters, or messages, give the underlying language model a template for vocabulary, tone, and emotional register. The model uses this context to generate responses that approximate how that person communicated.
Why was the product designed for short sessions rather than ongoing use?
The design philosophy was rooted in the concept of closure. Rocks compared Seance AI to a traditional seance: a bounded ritual, not an ongoing relationship. The team moved toward a pay-per-session pricing model specifically to discourage repeated use. A brief, meaningful interaction is the goal; dependency on a simulated version of a deceased person is not.
What did the premium tier include?
The premium tier offered voice recreation, where users could upload up to three audio samples of the deceased under five minutes each, which the system used to reconstruct a voice model. It also included animated images, which applied subtle eye and head movement to photos of the deceased. These features added a richer sensory dimension to the base text experience.
How did AE Studio approach the ethical dimensions of building grief tech?
The team was deliberate about transparency from the start. Rather than softening the premise, they named the product Seance AI and framed it openly as a fictionalized, AI-generated experience. Rocks publicly stated that AE Studio was greatly concerned about AI safety and wanted to draw attention to the implications of the technology, while supporting healthy regulation. The product itself was designed with explicit limits, short sessions and an honest acknowledgment that whatever the AI produced was not actually the deceased person.
Is Seance AI available to the public?
Yes. Seance AI launched in May 2023 with a free tier and a premium paid tier. The free tier allows users to conduct seances and store conversations for up to one month. The premium tier adds voice recreation and animated image features. The product is accessible at ae.studio/seanceai.

LET'S TALK

Bring us the hard problem.

We'll bring the team that ships.

Get in touch