A) Our Evaluation:
| Summary |
A stable diffusion web UI. Started as a free app, now has a subscription plan which includes 300 min of computational time. |
| Accessibility |
Requires some knowledge about how Stable Diffusion operates. |
| Limitations |
It is a stripped down version of other SD web UIs, which aims to appeal to the average Photoshop user. This is its main advantage but to some, this might be a shortcoming. Upscaling Images is not the best offered. |
| Specifics |
Stores input images and render results. Very good Inpainting functionality |
| Q&A |
Developer - Well documented website with regular contact options. A lot of tutorials on their website. |
|
Manuals - they call this section of their website “Using Invoke” (external link) |
|
Videos - They are regularly uploading Invoke videos on Youtube. Just search for Invoke, it is relatively popular as there is also a free version. |
| Last evaluation |
04 Feb 2025 |
| Status |
Used for two days for the Halle Competition. |
| Note … |
|
A “Fundamentals” video on InvokeAI
https://youtu.be/m8hnpX2XzL0?si=iCHwHV7kGr0ELIrs
B) This is what Perplexity says:
Invoke: Secure Generative AI Platform for Creative Teams-Custom Workflows, Model Training, and Full Ownership of Your Visual Media
Invoke is a professional generative AI platform designed for creative studios and teams to produce visual media securely and efficiently. It offers powerful creation tools, workflow builders, and collaboration features, allowing users to control every aspect of their creative process-from text or image prompts to finished assets. Invoke supports custom model training on proprietary content, upscaling, and integration with popular models like SDXL, Flux, and ChatGPT 4o. The platform emphasizes data ownership, security (SOC-2 compliant), and scalability, providing both cloud-based and downloadable community editions. Invoke is used across industries such as gaming, film, architecture, and product design to accelerate asset creation while maintaining full control over intellectual property.
SEG Tests
Archviz
SEG used Invoke as part of their work for the Halle competition. It was worked on two images : an exterior and an interior shot. Both images were started as an image to image workflow with ControlNet Depth map.
Emphasis was on obeying the existing geometry, while enhancing the lighting and the overall atmosphere of the image. In the examples below I will also provide the starting images.
-
Exterior Image of SEG’s competition entry for Halle Zukunftszentrum building and the adjacent park.
-
Starting image:
1.
-
InvokeAI end result:
-
Interior Image of the Lobby of the competition entry. This image was not part of the project presentation but presents better Invoke abilities. There are several not so straight lines that have been unaddressed, and of course a lot more entourage, furniture and vegetation can be added but the atmosphere of the space and its main architectural features are evident.
-
Starting Image:
-
InvokeAI End result
Concept exploration:
As in any other SD web ui, the creative freedom left to the AI can be restricted to different degrees. In the following images a little more breathing space was left to the AI:
Starting Image:
Results:
