Using AI to Make Composites : Firefly, MidJourney & Generative Fill in Photoshop Beta

The speaker begins by welcoming the audience to the video and announces the topic: using AI to create composite images of NFL football players. They discuss their personal history with photography and sports memorabilia, which led to their career as a photographer. The speaker shares their experience photographing NFL and NBA draft picks for trading cards and describes their unique approach to the photoshoots this year, which involved setting up five different lighting scenarios in a short time using the Ellen Chrome mobile app to control multiple lights. They also touch on their process of selecting and lighting the athletes, and how they used clamshell lighting and other techniques to achieve the desired look for the trading cards.

In this paragraph, the speaker delves into the post-production process, explaining how they used AI tools to retouch the athletes' skin and create backgrounds for the composite images. They discuss using vato AI for skin retouching, Firefly and mid-journey for generating backgrounds, and generative fill to fix any issues with the backgrounds. The speaker also mentions using Photoshop to blend all elements together and their approach to lighting the photos. They describe their process of selecting their favorite images from the shoot to use in their portfolio and how they used AI tools to enhance these images, focusing on learning and artistic expression rather than client requirements.

The speaker continues by detailing their process for creating the backgrounds using Adobe Firefly and mid-journey. They explain that Firefly was effective for generating cityscapes and landscapes, while mid-journey was better for stadium scenes and tunnels. The speaker shares the specific prompts they used in each AI tool to generate the desired background images. They also discuss their technique for removing the player from the generated background image using Photoshop's generative fill and then merging the layers. The paragraph concludes with the speaker's process for upscaling the AI-generated backgrounds using Photoshop's neural filters to improve their resolution and quality.

Here, the speaker describes the process of combining the retouched player images with the generated backgrounds in Photoshop. They explain how to use the remove background feature, refine the mask using the select and mask tool, and clean up the edges of the player's hair for a better selection. The speaker also discusses using neural filters to harmonize the color and contrast between the foreground and background layers, creating a more cohesive image. They further enhance the image by adding blue edge lights, adjusting blending modes, and adding grain to unify the image and give it a more realistic look.

In the final paragraph, the speaker adds the final touches to the composite image, including adding a cyan overlay, adjusting exposure, and painting in catch lights in the player's eyes to draw the viewer's attention. They explain their reasoning for not using generative fill for backgrounds, citing the time it takes and the limited options it provides compared to using Firefly and mid-journey. The speaker expresses excitement about the growing number of creative options available with AI tools and thanks the audience for watching. They encourage viewers to like, subscribe, and explore more of their content, emphasizing that they were not sponsored by any companies in this video and shared their learnings in the hope that others might find them useful.