PenMimic

  DOODLE IN. ART OUT. LIVE.

----------------------------------------------------------------

PenMimic is a tiny always-on-top window that watches any part of

your screen and repaints it with AI in near-real-time. Sketch in

ANY app you already love — Krita, Photoshop, MS Paint, even a

browser whiteboard — and watch the mimic turn your rough shapes

into finished-looking art as you draw, stroke by stroke.

No plugins. No cloud. No subscription. Everything runs locally

on your own GPU — your art never leaves your machine.

----------------------------------------------------------------

  HOW IT FEELS

----------------------------------------------------------------

  1. Drag a box over your canvas / 3d program / camera view of your drawing (whatever works) 

  2. Type what you're going for — "ink illustration of a

     dragon, moody lighting".

  3. Draw. Every change you make is re-rendered in well under

     a second.

Turn the strength knob down to keep your composition and just

polish it — or crank it up and let the AI hallucinate freely on

top of your scribbles. Lock a seed you love, or roll the dice

and get a fresh interpretation with every stroke.

----------------------------------------------------------------

  FEATURES

----------------------------------------------------------------

  *  Live img2img from any screen region — works with every

     drawing app / 3d app / anything on your screen that's under your capture area

  *  One-click setup: the app downloads its own AI engine and

     models. No Python, no command line, no fiddling

  *  Three built-in models from light to heavy (2 GB laptop

     GPUs and up), plus drop-in support for your own Stable

     Diffusion checkpoints and LoRAs (.safetensors / .ckpt,

     Civitai links built into the menus)

  *  Fully portable: the whole app — engine, models, outputs —

     lives in one folder. Put it on a thumb drive, plug it into

     another PC, it just runs

  *  Frictionless capture: click to copy a render to the

     clipboard, middle-click to save it in it's own /output folder inside the app folder, hover to pause

  *  Your knobs: prompt, strength, steps, resolution, seed +

     random mode, LoRA on top of any checkpoint

----------------------------------------------------------------

  REQUIREMENTS

----------------------------------------------------------------

  -  Windows 10 / 11 (64-bit)

  -  NVIDIA GPU — 2 GB VRAM minimum (DreamShaper 8 + LCM),

     7 GB+ for the best model (SDXL Turbo)

  -  ~3 GB one-time engine download, plus 2-7 GB for the model

     of your choice

  -  Download size of the app itself: ~300 MB

----------------------------------------------------------------

  HONEST SMALL PRINT

----------------------------------------------------------------

  -  First launch downloads the AI engine (~3 GB) — the pulsing

     green button walks you through it.

  -  Live speed depends on your GPU: roughly 0.6-0.9 s per

     frame on an RTX 3090 at 512px; slower cards should start

     with the smaller models.

  -  Windows may show a SmartScreen warning on first run (the

     app is unsigned) — choose "More info -> Run anyway".

================================================================

     Feed the mimic your doodles. It bites back beautifully.