Hello. Could Deep Art post a tutorial on how to provide optimal images for style creation when using the Abstract and Vincent AI’s? For example, should images provided for styles be of a certain size or aspect ratio? What characteristics of the image will result in good styles for applying to new images? Although some of the provided styles are quite evocative, I haven’t had much luck in creating similar globally good styles, they mostly tend to be rendered in a sort of rubber stamp repetitive way. Styles that create larger scale features seem difficult to create. Thanks!
Hi @IanJ. To get greater scale features set your chunk size setting to a higher value. Your style image should be at the minimum size of the chunk size. Further you can change the brush size setting to a value greater than one.
Thanks for this information. I’m not clear what is meant “tile size”. I don’t think you mean the size of the style sample, which I believe you call the “style image” in your email.
Under the controls for either the Abstract AI or the Vincent AI I see only two variables, “Brush Size” and “Style Weight,” but nothing about “tile size.”
It would be helpful to know what original images were used for each of the built in styles so they could serve as examples, but perhaps that’s not possible. Would it be true that if anyone supplied the same style image that was used for one of the built-in effects that the results of applying it, using either the Abstract or Vincent AI’s, would look the same as what is currently provided by the built-in styles? I guess what I’m asking is were the built-in styles created by the same software that you have provided as the Deep Art Effects program, or where they created by other means.
Thanks again!
Please go to File -> Preferences to find the chunk size setting. The built-in styles are speed optimized but based on the same Deep Art Effects core. Generally the Vincent AI works very well for custom styles based on textures and paintings. In the next version you will find a hint at each style with the name of the artist.
Thanks for this detailed information. Increasing the chunk size indeed does give a more “global” or larger-feature appearance of the supplied style to the final look.
One last clarification, please. In previous replies it was said: “Your style image should be at the minimum size of the chunk size.” I’m assuming first that “tile size” and “chunk size” mean the same. So, if I set the chunk size to 2048, then my style image should also be no less than 2048x2048 pixels, is that correct? If I do not size the style image, does Deep Art Effects auto-size it up or down into the appropriate range?
Thanks again for these helpful directions.
The AI will automatically scale up or down your style image based on the chunk size. For best quality you should use a style image with minimum size of the chunk size.