Analysis updated 2026-05-18
Generate a novelty Magic Eye style image from any text prompt for fun or sharing.
Study how a multi step AI pipeline chains prompt rewriting, image generation, and depth estimation together.
Run the project locally with mocked responses to explore the code with no API costs.
Browse the public Term Atlas to see previously generated stereogram terms.
| mikeville/text-to-stereogram | 0xkinno/astraea | 0xkinno/halcyon | |
|---|---|---|---|
| Stars | 0 | 0 | 0 |
| Language | TypeScript | TypeScript | TypeScript |
| Setup difficulty | easy | hard | hard |
| Complexity | 3/5 | 4/5 | 4/5 |
| Audience | vibe coder | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
Local dev mode uses mocked AI calls with no API keys needed, live providers require Anthropic and fal.ai keys.
This project turns a short text prompt into a Magic Eye style hidden image, the kind of picture where a 3D shape appears if you stare at a flat, repeating pattern in a certain way. You type in a word or phrase, and the tool generates a hidden subject, works out how far away each part of that subject should look, and builds a patterned surface around it, all from that one prompt. The author notes that most similar tools require you to supply your own picture first and just apply a preset pattern to it, while this one creates the subject itself. Under the hood the process runs in stages. First it rewrites the prompt into a clearer description and a matching pattern idea. Then it creates the subject image and the background pattern at the same time. After that it estimates depth, meaning which parts should appear closer or farther, using a model called Depth Anything that runs right in the browser. Finally it renders the finished stereogram image, adding small guide dots that help your eyes find the right focus point. By default, running the project locally uses fake, pre made responses instead of real AI services, so you do not need any API keys to try it out. If you want the real version, you can turn on live providers such as Anthropic for the prompt rewriting and fal.ai for the image generation, with Replicate available as a backup option. A public page called the Term Atlas shows past search terms that have already been generated and cached, so visitors can see completed examples without waiting for new ones. The project is built with Vite, React, and TypeScript, along with Netlify serverless functions for the parts that talk to outside AI services. It includes a debug panel, accessible by adding a special flag to the page address, that shows technical details like render settings and quality checks for people who want to dig deeper. To actually view the hidden image once generated, you look at the picture from about a foot away until two small dots at the top appear to become three.
A web tool that generates Magic Eye style hidden 3D images (autostereograms) from a plain text prompt, creating both the hidden subject and its pattern automatically.
Mainly TypeScript. The stack also includes TypeScript, React, Vite.
The README does not state a license, so usage terms are unclear.
Setup difficulty is rated easy, with roughly 5min to a first successful run.
Mainly vibe coder.
This repo across BitVibe Labs
Verify against the repo before relying on details.