Thaeyne, renowned for their insightful explorations of AI models, has once again delved into the world of technology, this time putting the spotlight on the Midjourney 5.2 inpainting model. In a recent video, they conducted comprehensive tests using both regular and raw modes, with a focus on 50 distinct words. The inpainting mode, known for its uniqueness, is not traditionally promptable. However, Thaeyne ingeniously approached this challenge by prompting the model in the usual manner and subsequently subjecting the resulting images to various region adjustments, covering the entirety of each image.
Throughout the experiment, Thaeyne experimented with a wide spectrum of over 300 words. The resulting video offers a tantalizing glimpse into the possibilities unlocked by the Midjourney 5.2 inpainting feature. By employing the technique of prompting the model conventionally and then applying a 'vary region' strategy to manipulate the images, Thaeyne managed to generate a diverse range of visuals, with a curated selection of 50 showcased in the captivating video demonstration.
The showcase of Midjourney 5.2's inpainting capabilities revealed a marked departure from the norm. The model exhibited images distinguished by superior lighting, enhanced color contrast, and a departure from the surreal and maximalist tendencies commonly associated with the regular model. Notably, the inpainting model displayed a penchant for producing shades of red, diverging from the orange tones typically seen in the regular model's output.
Perhaps the most intriguing revelation to emerge from this experiment was the substantial disparity between the images produced by the inpainting model and the original inputs. This divergence underscores the model's remarkable capacity to generate unique and imaginative visuals, largely unaffected by the initial image, highlighting its creativity and innovation.
However, it's worth noting that the Midjourney 5.2 inpainting model is currently not available for standalone prompting. Nevertheless, there are promising plans on the horizon to incorporate this functionality into a future version. This development is eagerly awaited by enthusiasts and creators alike, as it will open up new avenues for exploring and harnessing the potential of this groundbreaking AI technology.