Invoke AI recently launched two new features on its AI-based image creation platform. According to the company, these two new features - Model Trainer and Control Layers - provide unprecedented fine control for image generation. Both tools empower users to control how AI creates and modifies images in a highly refined manner. In addition, Invoke announced that it has obtained SOC 2 compliance certification, marking the company's high level of data security.

In a conversation with GamesBeat, Invoke's CEO Kent Keirsey detailed these new features and discussed how they provide greater control and refinement for image creation. The Model Trainer is particularly customized, allowing businesses to train specific image generation models using just 12 pieces of their own content. Keirsey stated that this customized training can produce more coherent images that match the developer's IP, enabling AI to create artworks with the same style and design features more frequently.

"We are helping the models understand what we mean when we use certain language," explained Keirsey. "When we explicitly ask for a certain interpretation, it means we need 10 to 20 images that reflect that idea, that style, to train... We're saying, 'This is our studio style, and it involves different themes.' Whether it's for a general artistic style or specific intellectual property, this approach can be adopted."

Invoke emphasizes that one of its goals is to provide higher security, which is why they obtained SOC 2 compliance certification. This high level of security reduces the risk of developers' images being used to assist in creating other studio intellectual properties.

Regarding how to train AI, Keirsey further demonstrated the Control Layers feature. This feature allows users to segment specific areas of an image and assign specific prompts to those areas. For example, users can use the layer tool to draw in the upper corner of an image and give AI a prompt to place a celestial body in that specific location. This way, creators can adjust the composition of the image and control individual parts without having to change the entire image.

The prompts attached to each layer can be refined and generated like any other AI image, but these effects only apply to the specified parts of the image. Additionally, Control Layers allow users to upload specific content to layers, where creators can choose what specific elements they want AI to retain from the image, such as style, composition, color, etc.

When asked about how Invoke's new tools integrate into the game development workflow, Keirsey stated that many developers are cautious about using AI, and one important reason is copyright issues. "The human concept has to exist - human sketches, the initial human idea. Developers will draw a line and say, 'None of this is going into the game. Unless we can prove we can get the copyright, we're not going to take the risk.' Once you can prove copyright ownership, you'll see AI-generated content entering the game... That's why Invoke is trying to answer this question for organizations, showing human expression, giving them more ways to showcase, so that we can prove copyright and expedite this process."