Documenting process is important. I'm glad we have space on SuperRare now to share these deeper thoughts. This is a repost from my Patreon and just wanted it here cause I would rather have all my writings be paired with my works ultimately:
How do I blend my painting with Midjourney to create these works? People have been asking, so I’m writing where the experiments are at for those interested and to document the process.
Most of my Midjourney generations start from my oil paintings moodboard. That board is wide and detailed photos of oils I’ve made, shot to catch how I layer paint and how my brushstrokes sit. I keep other moodboards too: digital paintings, watercolors, and styles I’ve built. I mix those into the oil set when I want something else in the mix. I also have a long trail of image ratings feeding my personalization profile. The oil board is still my favorite dataset.
For the self portrait series especially, I sometimes use a photo as an image prompt with text prompts, then refine through prompting. That path is faster. Sometimes I only prompt with my style customizations and see what comes out with no photo holding the pose.
Now that I’ve been mending from injuries, I take a generation into Procreate and work on it there. I reinforce my brushstrokes and put my hand back into the piece. That step is relaxing because the Midjourney generation is already like a rough sketch painting built from my painting data and prompts, so I can focus on refining forms and blending color, which is my favorite part of painting.
Timelapse Video:
Then I take that refined work and put it back into Midjourney as an image prompt. The back and forth can run as many or as few times as I want until the result satisfies me.
I like collage elements in the finals, glitches, and some of the AI aberrations. I’m not trying to hide the AI or make every inch look like a classical painting where every mark is under my control. I’m trying to see what visual language shows up when I collaborate with the machine.
Almost everything after the original oils can be done on an iPad or tablet. That keeps the process accessible compared with local open source models that need more install and upkeep, which is why I lean on Midjourney or Grok. For now Midjourney still wins on stylization and consistency. Grok is less censored and better at video, so I use it for figures and animation. Midjourney has loosened some of the word limits around adult human figures, which helps. I’ve been on Midjourney since v2, with a protest hiatus for a few months. I came back when they announced medical imaging work in the real world, and when some of that language control eased.
That’s about it. That’s the slop process.
🖤
PS ~ Thank you @cozomo_de_medici for adding this work to your @mediciemerging23 collection.
Comments (14)
Love this empress!
Thank you @kaelivanfossen
🔥🌈
Fascinating. Thanks for sharing your process. The more I know about true artists using AI the better my contributions to the conversation.
Interesting process and overview of it.