Keep ideas, assets, and outputs in one workspace.
A local GPT-image-2 workflow for designers. References, prompt assets, concurrent tasks, and history stay together.
Prompts should not be rebuilt from scratch.
Turn references, color constraints, and long prompt fragments into reusable creative atoms.
Turn recurring people, products, and styles into a visual library.
Assets do not end with one task. Public Gallery and the image editor carry them into the next round.
Save uploads, recent references, or generated results, then organize them by people, products, and styles.
Add names, usage notes, and reference notes. Replace, delete, or drag to reorder at any time.
Crop, combine layers, transform proportions, erase locally, paint, and add directional marks on device.
Select, stage, add as a reference, download, or archive generated results.
Keep creating. Every result stays findable.
Work in the generation view, then recover results by keyword, month, mode, quality, ratio, or orientation.
Concurrent tasks enter a clear queue. Retry partial failures without repeating the entire generation.
Local thumbnails, cursor pagination, and archives keep a growing task library responsive.
OpenAI-compatible API, local Codex mode, and CLI are available as secondary paths beyond the visual workflow.
Choose the build that fits this machine.
Standard packages are recommended for new users. Every link goes directly to GitHub Releases.
Current version 0.6.1Read the full release notes
Standard packages
Show all three portable downloads
macOS builds are currently unsigned and not notarized. If launch is blocked, right-click the App and choose Open.
Standard builds use the system application data directory. Portable builds keep data in a sibling data directory.
The project uses AGPL-3.0-only. Advanced local Codex mode depends on local authorization; API mode is better for stable or shared setups.