InvokeAI 2.1 has been released. Now you can redraw parts of images, easily replace objects with preserving the context (cats for dogs, blue hair for green) and add any other models besides SD, on which the style of images depends — there are already hundreds of them.
The SD model itself has been updated to version 1.5, it's a little bit more trained — there are less deformed faces and hands with six fingers (we have two separate neural networks for quick correction of faces, by the way).
I am very happy with the pace of development of the project: 2 months ago we had only the command line for generation, and now we have the best GUI for today. I'm praising myself here, yes, but I’m really proud.
We're incredibly lucky to have top programmers who are experts in computer graphics and neural networks. We have a great frontend as far as we can at our pace — developers joke that we spend more time on one button than others do on the entire UI. Most of my suggestions did get implemented and I'm extremely pleased that I can now use the interface myself (I've never said «I told you so!» so often).
In fact, there is still a lot of work to be done and new ideas and features appear every day.
GitHub https://github.com/invoke-ai/InvokeAI/
Discord https://discord.gg/ZmtBAhwWhy
Video Demo
- Inpainting: Inpainting is live in the WebUI! A robust inpainting canvas and controls system has been added for use in the browser app.
- Cross Attention Control & Prompt Syntax Updates: A novel and easy-to-use prompt syntax, including weights, swaps, blends, and more, is now baked into the prompting system in the WebUI and CLI.
- Model Switching: Easily import, switch and load models from the CLI and WebUI.
- UI & Workflow Enhancements: Aside from a number of small UI enhancements, the Gallery and options panel have received massive overhauls, allowing for dynamic control of your canvas and workspace.
- (Beta) 1 Click Installer: Getting started with InvokeAI has never been easier — We’re releasing a beta version of our 1-click installation bundles to get everything set up for you, and guide through every step of the process including model downloads.
The pace of development and excitement in the community continues to grow, with everyone from professional game development studios to artists exploring ways to use the power of InvokeAI to accelerate their creativity. We've got a lot planned, with Outpainting coming soon, more model management features, and a novel backend implementation that will set the standard for architectural excellence.
Whether you're a dev looking to build on or contribute to the project, a professional looking for pro-grade tools to incorporate into your workflow, or just looking for the best open-source SD experience out there, we're looking forward to you joining the community.
You can get the latest version on Github, and join the community's Discord to participate in exploring Invoke.
• • •