LukaDev13/Liminal-Dreamcore-1K

Dreamcore

A Collection of 1000 AI-Generated Dreamcore Aesthetic Images

---

All images in this collection are AI-generated.

---

Architecture

The generation pipeline behind this collection:

---

What Is Dreamcore?

Dreamcore is an internet aesthetic that captures the visual language of dreams - specifically the strange, liminal, half-remembered quality of dream imagery. It sits in the same family as weirdcore, traumacore, and oddcore, but has its own distinct visual identity.

Dreamcore images typically feature:

• Liminal spaces - empty playgrounds, vacant hallways, abandoned pools, places that feel transitional and uninhabited
• Nostalgic childhood elements - bright colors, cartoonish suns with faces, toy-like objects, playground equipment
• Surreal juxtapositions - things that do not belong together placed in the same frame, creating an uncanny feeling
• Soft, hazy lighting - overexposed skies, bloom effects, a washed-out quality that mimics how dreams feel visually
• Familiar yet wrong - scenes that look like places you have been before, but something is off. The geometry is slightly wrong. The colors are too saturated or too pale. There is a presence you cannot identify.
• Emotional ambiguity - dreamcore images can feel simultaneously comforting and unsettling. They evoke the feeling of a childhood memory you cannot quite place, or a dream you woke up from and tried to hold onto.

The aesthetic draws heavily from early 2000s internet imagery, educational materials, children's media, and the visual texture of places that exist between states of use - not quite abandoned, not quite alive. It is the visual equivalent of deja vu.

Demo Images

A small sample from the 1000-image collection:

004.jpg	086.jpg	089.jpg
102.jpg	261.jpg	278.jpg

How These Images Were Generated

All 1000 images in this collection were generated using GPT Image 2 at 2K resolution with medium quality settings.

Generation Pipeline

Each image was produced through a structured prompting process:

1. System prompt - A carefully constructed system prompt defines the dreamcore aesthetic parameters, ensuring consistency across the entire collection
2. Image model - GPT Image 2 is used as the generation engine
3. Resolution - 2K (2048x2048 or equivalent) for sufficient detail and clarity
4. Quality setting - Medium, balancing generation speed with visual fidelity
5. Sequential numbering - Images are numbered 001.jpg through 1000.jpg for easy reference

The system prompt used for generation enforced the core dreamcore visual elements: liminal spaces, nostalgic childhood imagery, surreal compositions, soft lighting, and the characteristic uncanny familiarity that defines the aesthetic.

Collection Structure

Dreamcore/
├── 001.jpg
├── 002.jpg
├── ...
├── 999.jpg
├── 1000.jpg
├── README.md
└── LICENSE

All 1000 images are stored as JPEG files with zero-padded sequential numbering. The naming convention makes it easy to reference specific images or iterate through the collection programmatically.

License

This project is licensed under the MIT License. See the LICENSE file for details.

All images are AI-generated and released under the terms of the MIT License. You are free to use, modify, and distribute them for any purpose, including commercial use, but you must credit the author (Luka).

---

1000 images. One aesthetic. Every single image is AI-generated.

github.com/Luka12-dev