Help Keep Lilith AI Running and Growing

Hi, I’m the person behind Lilith AI.

For the past few months, I have been building an AI model inspired by The NOexistenceN of you AND me. This is not a generic chatbot. Lilith AI is trained directly on in-game dialogue and story material and taught to be Lilith, capturing the personality and tone of the original character. I'm also a full-time student, so this project is something I work on in my own time.

The project has been publicly available for about three months, with major improvements in the last few weeks. You can try it now at this website on desktop or mobile without downloading anything. The model is also available for download and use in LM Studio or Jan, though vision features are not supported in that version.

---

What Makes Lilith AI Different

- Privacy-focused: chats are stored on your device through the browser. The only information tracked is total users and total messages sent.

- Uncensored: no unnecessary filters.

- Ad-free: I want to avoid ads whenever possible.

- Character-driven: Lilith AI is trained to embody a specific character, not act like a generic assistant.

This is a small, independent project. I built the software and model, and a few close friends helped set up the server.

---

Current Usage

Despite being a small project, Lilith AI has reached about 2,000+ users, processed 5,000+ messages, and has 8000+ downloads! Feedback and discussion are available on Reddit for anyone interested in seeing what people think.

---

Why I Need Help

The AI runs on a limited server:

- NVIDIA TITAN X (12GB VRAM) - around GTX 1080 Ti performance

- NVIDIA GeForce GTA 1080 (8GB)

- 32GB DDR5 RAM

It works, but it has limitations:

- Responses can be slow

- Context length and memory are limited

- Weekly maintenance is required

- Some planned features cannot run

Upgrading the hardware would immediately improve response times, allow much larger context length, and reduce downtime.

---

Planned Features with Upgraded Hardware

After the hardware is upgraded, I plan to add:

Encrypted accounts: users can sync chats across devices without compromising privacy. No data will be sold or used for training without consent.

End-branching mode: because the model is trained on The NOexistenceN of you AND me, users will be able to pick different endings (good, neutral, bad), and the AI will be filled with all story events for that ending. This allows talking to Lilith within that version of the world.

Other community-requested improvements and vision upgrades will also follow.

---

If Funding Does Not Happen

I will continue to run Lilith AI myself for as long as I can, but the hardware limitations will prevent encrypted accounts, end-branching mode, and major context improvements. The AI will still be available, but growth will be limited.

---

Why I Built This

I love experimenting with AI, and The NOexistenceN of you AND me had a strong impact on me. Lilith AI is a combination of that experience and my own ideals around privacy, freedom, and creativity.

If you enjoy Lilith AI or want to support an independent, privacy-focused project, even a small donation makes a real difference. Thank you for helping make this project possible.