You'll likely be familiar with Image Generation in general but here is a brief recap as well as an introduction to working with 3D files.
Starting with 2D generation we are going to open the Image Generation tool and start by choosing our AI Model. Here I am using Imagen and Stable Diffusion
, but you can choose from any that are available.
Moving down to the Compose tool you'll find a canvas and a lighting slider, which I am going to increase to 75%. I am going to use one of Pencil's sample packshots but you can of course use any image you wish including those in your Asset library.
I have positioned the asset centrally in the canvas and now I will move to the Prompting tool where I can prompt for the scene I want. I am referring to the packshot as the asset rather than a glass of beer so that the AI model doesn't inadvertently generate an additional glass of beer.
I am going to set my aspect ratio as 1:1 and choose to generate five images at a time.
The five images generated by Stable Diffusion I feel are a better response to the prompt, clearly showing reflecting sunlight and some wooden, rural bars in the background.
The benefit of being able to generate five or ten images at a time is that you have a decent chance of generating something you want to work with. It's really critical to have packshots from multiple angles so that you can prompt for a variety of different scenes. If you don't have packshots from a variety of angles you can make use of our 3D Composition tool.
Going back to the Image generation tool, I am going to slide to the 3D option at the top of the Compose tool and upload a USDZ file. We currently support USDZ and GLB files and our 3D file requirements are here. I am going to upload an image of a sneaker/sports shoe.
I can move the asset around the canvas as a 3D object. I am going to set the lighting slider to 75% again, seeking good integration with the prompted scene.
There are then some Advanced 3D Settings to take a look at:
HDR: There are some preset HDR modes you can choose from as well as a HDR Intensity slider you can use to directly adjust the amount of HDR effect applied to the image.
Shadow: Here you can apply a shadow by defining the light direction in degrees.
Again, in my prompt I am going to refer to the sneaker as the asset to mitigate against any additional sneakers being generated.
Overall, I am really pleased with the results, particularly the lighting and the integration of the sneaker into the scene. One criticism would be that they are all fairly similar but this could be improved with more detailed prompting and/or the use of different models.