jev-paint
Local browser app that turns Jev's per-pixel colour probability distributions into textured paintings.
Overview
jev-paint is a small local creative experiment that paints images from Jev's probability distributions. You choose a representation (palette, HSL, binary RGB or silhouette) and grid size and enter a prompt; prompts describe pixels and regions rather than sending raw images. The browser builds typed questions per pixel, sends them in four concurrent batches, and turns the returned distributions into a painting.
The renderer is deterministic: spatially correlated categorical sampling supplies stroke pigments, the mean colour supplies the underpainting, entropy controls relief, strokes follow luminance contours and a height map adds impasto lighting. Per the list summary, predicted confidence shapes how strokes are drawn. HSL and RGB channels are treated as independent because Jev gives no joint distribution. A Python standard-library helper serves files and forwards requests, since the API disallows browser CORS. The README stresses this is an artistic transformation, not a probability chart. Offline tests use recorded fixtures.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | What colour (palette entry or channel level) is this pixel for the prompt?(inferred) | — |
How it works
- Start the local helper and paste a Jev key
- Choose representation, grid size and prompt
- Send per-pixel typed questions in batches
- Render the distributions into a painting in a worker
Strengths
- No packages or build step
- Offline tests and benchmarks with recorded fixtures
- Key and prompts sent only to TypeSafe
Caveats
- Larger grids mean more paid requests
- Key stored unencrypted in browser localStorage
- Artistic output, not a faithful image
Getting started
python3 server.py, then open http://127.0.0.1:8791 and paste your Jev key in Settings.At a glance
- Type
- Browser extension
- Status
- prototype
- Language
- JavaScript
- Runtime
- Browser app with Python helper
- Has code
- yes
- Has tests
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
Creative experiment: paints images by having Jev predict every pixel's colour in parallel, with predicted confidence deciding how wide each stroke is drawn.
yibie/awesome-jevRequires Python 3.9+ and a modern browser with module workers and OffscreenCanvas (current Chrome, Edge, Firefox, or Safari). No packages, build step, or Node installation needed.
logicrw/awesome-jev-projectsLocal browser app that turns batched per-pixel Jev distributions into paintings; prompts describe pixels and regions rather than sending raw images.
cobanov/awesome-jevLocal app that turns Jev's per-pixel probability distributions into paintings.
MrJev/awesome-jev
Frequently asked questions
How does Jev paint?
It gives a colour probability distribution for each pixel, which a deterministic renderer turns into strokes.
Does it send images to Jev?
No. Prompts describe pixels and regions in text.
Do I need Node?
No. Python 3.9+ and a modern browser are enough; Node is only used for tests.
Related projects
Toy chat that makes Jev talk by repeatedly choosing its next word from a tiny caveman vocabulary.
ProjectprototypeTypeScriptText generation with Jev via cascaded Choice questions: word type, WordNet category, embedding cluster, then the word itself.
ProjectresearchPythonExperiment that uses Jev as a sampler: DeepSeek proposes top-k next tokens and Jev chooses which comes next and when to stop.
ProjectresearchPythonExperiment that makes Jev produce prose by asking one choice question per word over a routed domain vocabulary.
LibraryresearchJavaScriptPlayful web arena where Jev does arithmetic, writes chat replies and draws pixel art one Choice or Noul decision at a time.
DemoprototypeChatGPT-style web chat where every word of every reply is a Jev multiple-choice decision over a ~1,800-word vocabulary plus END.
ProjectprototypeTypeScript