AfterWords

Inspiration

AfterWords was inspired by a simple but painful problem: people often lose someone with words still left unsaid.

That loss can come from death, a breakup, distance, or any relationship that ended without closure. In those moments, text alone can feel cold. Voice carries emotion, memory, and familiarity in a way that normal chat cannot.

We built AfterWords to use voice AI for something deeply human: helping people process grief, heartbreak, and unresolved goodbyes through one final conversation that feels personal, familiar, and emotionally meaningful.

What it does

AfterWords lets a user upload real chat history and a voice recording from someone they miss.

The app studies how that person communicated, including their tone, common phrases, message length, emojis, and emotional style. It also uses the uploaded audio to recreate a familiar voice.

Then the user can have a conversation in a simple chat interface. The AI responds in a style based on the real messages, and the response is played back in the person’s voice.

There is also a coping mode. When the user needs more direct support, they can switch modes and the app becomes more comforting and support-focused instead of only trying to match the person’s style.

Impact

AfterWords solves a real emotional problem. Many people deal with grief, heartbreak, or unresolved loss, and they often carry thoughts they never got to share.

The impact comes from voice. Hearing a familiar voice can make the experience feel more real, personal, and emotionally powerful than a normal chatbot. AfterWords gives users a private space to reflect, say what they need to say, and move toward closure.

The project is not trying to replace a real person or pretend someone is alive. It is designed as a support tool for reflection, comfort, and emotional processing.

Execution

We built a working full-stack prototype.

The user can upload a chat file and an audio file. The backend parses the chat, extracts the other person’s messages, builds a persona from their writing style, and connects that to an AI model for responses.

We use the Gemini API to generate emotionally appropriate replies based on the created persona. We use ElevenLabs for voice cloning and text-to-speech, so each response can be heard in a familiar voice.

The frontend displays the conversation in a clean chat interface and plays the generated voice response. The demo clearly shows the full flow: upload, persona creation, chat response, voice playback, and coping mode.

Creativity

AfterWords is different from a normal chatbot because it is built around voice as the emotional layer.

The app does not just generate text. It combines real chat history, persona generation, voice cloning, and text-to-speech to create a more personal experience. The familiar voice is what makes the demo memorable.

The creative part is using voice AI not just as a feature, but as the main emotional bridge between the user and the conversation.

Track fit: Best Project Using Voice AI

AfterWords strongly fits the Best Project Using Voice AI track because voice is central to the product.

The core experience depends on taking an uploaded voice recording, creating a reusable voice model, generating a response, and playing it back as speech. Without voice AI, AfterWords would just be another chatbot. With voice AI, it becomes a much more personal and emotionally meaningful experience.

We use voice cloning and text-to-speech to make the interaction feel familiar, human, and memorable. This is exactly where voice AI has the most impact: creating an experience that text alone cannot provide.

Challenges we faced

The biggest challenge was making the experience emotional without making it feel uncomfortable.

We had to be careful with the product framing. AfterWords is not about bringing someone back. It is about helping the user process their emotions and say what they need to say.

Technically, the hardest parts were connecting the chat parser, persona builder, Gemini response system, and ElevenLabs voice generation into one smooth demo. We also had to keep the project simple enough to finish during the hackathon, so we used an in-memory session system instead of spending time on accounts or a database.

What we learned

We learned that voice changes how people experience AI. The same sentence feels completely different when it is heard in a familiar voice instead of just read as text.

We also learned that strong AI projects need more than technical features. They need a real problem, a clear user, and a product experience that people understand immediately.

AfterWords taught us how to combine chat parsing, persona prompting, Gemini, voice cloning, text-to-speech, and emotional support into one meaningful voice AI experience.