mesh-avatar-studio
일러스트를 코딩 에이전트와 로컬 편집기로 2D 메쉬 애니메이션 아바타로 변환
디자이너·사진작가·AI 업스케일링 사용자를 위한 무료 오픈소스 이미지 비교 도구로, 동기화 뷰·고급 확대·WYSIWYG 내보내기를 지원합니다
A free, open-source tool for detailed visual image comparison. Ideal for designers, photographers, and AI upscaling enthusiasts. Features synchronized views, an advanced magnifier, and WYSIWYG export.
디자인Python · GLSLGitHub @Loganavter온라인운영자 미확인
써 보셨어요? 한 번만 눌러 주세요. 로그인은 필요 없어요.
Improved free before/after slider albums.
저장소 README에서 가져옴
README 전체를 불러오는 중입니다. 아래는 저장된 발췌입니다.
An open-source workspace platform for advanced image and media comparison. Read this in other languages: Русский • Get Improve-ImgSLI: Windows installer • Flathub • AUR • Install & run from source: docs/INSTALL.md • launcher reference: docs/LAUNCHER.md • Learn the app: in-app Find Action ( Ctrl+Shift+P ) and Help • browse topics: EN Getting Started • EN host topics • RU • Contribute: CONTRIBUTING.md • More: Development History • Project VisionREADME 전문 보기 ↗
BlinkTable opens on a session picker and gives you independent workspace tabs: a two-image Image Compare, a synchronized N-image Multi Compare, and an AI Gallery for triaging, measuring and AI-scoring whole folders. Several workspaces stay open at once, and each tab keeps its own images, zoom and help.
For retouching, upscaling QA and photography, where the comparison itself is what you deliver. It opens images at enormous megapixel resolutions without lag.
Below, the app's individual tabs.
Image Compare puts two images side by side behind one divider. Magnifiers are separate objects: several at once, each with its own area, zoom and colours. A comparison can be saved as a full-resolution image, or recorded in the app and trimmed in the built-in mini video editor.
AI Gallery does a lot of the work of comparing large series for you, quickly: local algorithms or an AI model over OpenRouter judge the shots, and the interface lets you move through the gallery without waiting.
Multi Compare is a convenient grid for comparing more than two images at once, dozens and up to hundreds, with one shared zoom and pan for the whole grid and slots you can drag.
An AI agent can drive a running window and read what is in it: open a tab, take
a comparison apart, capture a surface. Start the app with BLINKTABLE_CONTROL=1
so it can connect, then ask in plain words.
Reference: docs/dev/MCP.md.
| Platform | How |
|---|---|
| Windows | Installer from Releases |
| Linux (AUR) | yay -S blinktable |
| Linux (Flatpak) | flatpak install io.github.Loganavter.Improve-ImgSLI, then flatpak run io.github.Loganavter.Improve-ImgSLI |
| From source | git clone https://github.com/Loganavter/BlinkTable.git && cd BlinkTable && ./launcher.sh run |
Details: docs/INSTALL.md, docs/LAUNCHER.md.
Images above 65536 px on a side open only when the format can be read in a
stream; otherwise the decoder stops at that size.
| Topic | Where |
|---|---|
| Install, launcher | docs/INSTALL.md, docs/LAUNCHER.md |
| Controlling the app from an agent | docs/dev/MCP.md |
| Development setup | CONTRIBUTING.md |
| Architecture and contracts | docs/dev/ARCHITECTURE.md, docs/dev/CONTRACTS.md |
| Vision and history | VISION.md, docs/DEVELOPMENT_HISTORY.md |
| Open engineering debt | docs/dev/TODO.md |
Setup, coding guidelines and packaging notes are in CONTRIBUTING.md. Issues and pull requests go through GitHub.
GPL-3.0-or-later, see LICENSE. Runtime dependencies, including PySide6/Qt under LGPL-3.0-or-later, are listed in THIRD_PARTY_LICENSES.md.