One-pixel automation with instant stop control.
PixelLoop is a Windows desktop automation tool for users who need to trigger a fixed action loop based on a single screen pixel. It continuously checks one configured pixel color, clicks a defined location, performs a mouse scroll, counts loops, and stops immediately when the pixel changes. The product is for users who want a lightweight, deterministic utility with no OCR, no image recognition, and no window detection.
Power User Lucas, 29 - Lucas uses a Windows PC and wants a lightweight utility that performs the same action repeatedly while a screen condition remains true. He values deterministic behavior and low overhead.
Automation Tester Camille, 34 - Camille wants to validate that a repeated mouse action behaves exactly the same every time and that counters reflect the real number of iterations.
Casual User Mehdi, 41 - Mehdi is not technical and needs a polished Windows app with a minimal interface. He wants a simple start and stop workflow with visible results.
Lucas opens PixelLoop and the app instantly loads his config.json. He clicks Commencer and watches Boucles and Captures update as the tool repeats the exact click-and-scroll sequence whenever the target pixel stays orange.
A few minutes later, the screen state changes and the pixel no longer matches. PixelLoop stops immediately, freezes the final totals on screen, and Lucas can see exactly how many iterations completed without having to guess.
That determinism matters: he gets a tiny Windows utility that does one thing reliably, with no OCR, no image matching, and no complicated setup. The result is faster execution, fewer mistakes, and a tool that is easy to package and distribute as an .exe.
Team & resourcing - Small team - 1 Python engineer, 1 UI designer, part-time QA, and part-time PM.
Paste this into Cursor, Bolt, Lovable, or v0 to start building.
Build a Windows desktop automation app in Python called PixelLoop using PySide6. Goal: Create a local-only .exe-ready utility that continuously checks one screen pixel, and while that pixel matches the target color within a configurable tolerance, it clicks a fixed location, waits, scrolls, waits, increments a loop counter, and repeats. Stop immediately when the pixel no longer matches or when the user presses Arrêter. Hard requirements: Use only single-pixel color detection. Do not use OCR, image recognition, OpenCV template matching, window detection, or any game-specific logic beyond the pixel check and input actions. Files: main.py ui.py bot.py config.json Config: Load config.json automatically at startup. Required keys: pixel_x, pixel_y, click_x, click_y, scroll, delay, tolerance, multiplier Use target color #FF6600 in code. Delay is in milliseconds. Tolerance is numeric and user-configurable. Multiplier is used to compute Captures = Boucles × multiplier. UI: Modern PySide6 interface with only these visible controls and outputs: Commencer button Arrêter button Boucles label with live loop count Captures label with live derived count Show running/stopped and config validation state clearly. Behavior: When Commencer is pressed, begin a loop that samples the configured pixel continuously. If the pixel matches #FF6600 within tolerance, do this exact sequence: click at click_x, click_y wait delay ms scroll by scroll wait delay ms increment Boucles by 1 repeat If the pixel no longer matches, stop immediately and keep the final counts displayed. Architecture: main.py launches the app and wires UI to bot logic. ui.py contains the PySide6 window and widgets. bot.py contains the pixel reading, loop controller, action executor, config loading, validation, and counters. config.json contains default values and is read on startup. Implementation details: Use efficient single-pixel screen sampling suitable for Windows. Use a background worker thread or Qt thread so the UI stays responsive. Expose methods to start, stop, reset counts, and update live counters. Handle invalid config cleanly with user-facing error messages. Add structured logging to a local file for startup and runtime issues. Packaging: Prepare the project to be bundled with PyInstaller into a single Windows .exe. Avoid external cloud dependencies. Deliver the complete project file by file, starting with main.py, then wait for confirmation before continuing.
Je veux développer un logiciel Windows (.exe) en Python. Le logiciel est destiné à automatiser une manipulation dans Dofus. IMPORTANT : Je ne veux PAS utiliser d'OCR. Je ne veux PAS de reconnaissance d'image. Je ne veux PAS de détection de fenêtre. Je ne veux PAS changer la logique. Je veux uniquement reproduire exactement cette logique : 1. Vérifier en permanence le pixel : X = 490 Y = 592 2. Si la couleur du pixel est #FF6600 (avec une tolérance réglable) : - Cliquer à : X = 522 Y = 671 - Attendre 100 ms - Faire un scroll de souris de -120 - Attendre 100 ms - Ajouter 1 au compteur - Recommencer la boucle 3. Si le pixel n'est plus de la couleur #FF6600 : - Arrêter immédiatement la boucle - Afficher : Boucles = compteur Captures = compteur × 7 Je veux une interface graphique moderne avec PySide6. L'interface doit contenir : - un bouton "Commencer" - un bouton "Arrêter" - un label "Boucles" - un label "Captures" Je veux également un fichier config.json contenant : - pixel_x - pixel_y - click_x - click_y - scroll - delay - tolerance - multiplier Le programme doit lire automatiquement config.json au démarrage. Le logiciel doit être organisé proprement avec plusieurs fichiers : main.py ui.py bot.py config.json Je veux un projet propre, prêt à être compilé avec PyInstaller afin d'obtenir un .exe. IMPORTANT : Ne me donne pas des morceaux de code. Je veux le projet complet fichier par fichier. Attends ma validation avant de passer au fichier suivant. Pixel de détection X = 490 Y = 592 Couleur = #FF6600 Clique ici X = 522 Y = 671 Scroll = -120 Delay = 100 ms Tolérance = 30 Résultat = Nombre de boucles × 7 Si tu proposes une autre logique (OCR, OpenCV, reconnaissance d'image, recherche de fenêtre, etc.), refuse cette idée et reste strictement sur ma logique de détection d'un seul pixel.
Design by The Resonance | Powered by GPC – The AI Transformation Company