Dercius

Well, Black Friday is approaching. To celebrate it , I will be offering a discount on all works until the end of next week. 😋

Now would be the time if you are interested in a specific piece. Think about it again.

Now from GMT 12am till Friday GMT 11pm

Stay healthy.

Now with a larger audience and other fellow artists at hand, I would like to start conducting some basic surveys on specific topics. The first survey concerns the resolution of works.

The resolutions I currently use most often are 9:16, 1:1 and 2:1. 9:16 is the dominant resolution due to its widespread use on other platforms.

Furthermore, it would be interesting to hear what the preferred upscaled resolutions for works would be. 2k, 4k or others?

The following section goes into a little more technical detail. You can skip it if you are not interested.

Depending on the tool used, there may be limitations on the part of the software or hardware. I myself use Stable Diffusion (SD) locally in combination with Midjourney (MJ).

SD has hardware limitations and is quite flexible in terms of resolution and upscaling of works.

The resolutions I use most often are 9:16, 1:1 and 2:1. But anything is possible here.

Midjourney has presets of 1:2, 9:16, 3:4, 5:6, 1:1, 6:5, 4:3, 16:9 and 2:1. Manual resolution settings are also possible here. MJ is currently still poor at upscaling. However, this is not a problem which can be fixed with SD and other tools.

My experience has shown that there is a clear correlation between the resolution and the tool's interpretation of how it would attempt to create the work. This is particularly true when it comes to character representation.

This applies to close-ups as well as full-body portraits. This applies to SD and MJ.

SD has a solid understanding of portraits and the definition of close-ups to full-body shots. MJ has problems with this, which can be frustrating at times. But SD can also have problems with this, depending on how the checkpoint was trained.

Another observation is that the quality and detail improves between 1:2, 9:16, 3:4, 5:6 and 1:1. It decreases from 3:4, 5:6, 1:1, 6:5, 4:3, 16:9 to 2:1. This is more of an issue with MJ than with SD. The main problem here is that when the tool is used on a larger canvas, it tries to fill it with something, and this may be something that does not fit the desired vision. That is why 1:2, 9:16, 3:4, 5:6, 1:1 are also more able ot stick to the prompt.

Here the link to the poll:

https://www.deviantart.com/dercius/poll/Survey-8211-Preferred-resolution-8687300

Looking forward to your feedback and participation on this topic.