Hey everyone, it's been a while since I wrote even though my goal this year was to write more consistently. I rescued a puppy who was abandoned on my street (my gutter potato) in Feb which has been a lot with also having a rescue fighting dog, so writing got put on the back burner for a few cause trying to implement a new habit with puppy chaos was really too hard haha.

Now he's older and more trained and doesn't need my 24/7 care, I can do things like write interrupted with focus and expand again incorporating hopefully regularly writing into my rituals. I already write a lot on X, farcaster and zora with notes, so a new process I'm trying is to send all the links of my posts with notes to Grok to ease some of the initital starting pains of writing.

With Grok 4, I told it to review all my X posts and previous paragraph posts to learn my tone and writing style. Then I sent it all the different posts about my Liquid Glitch experiments, Zora x Tik Tok posting habits to write into an article. After a few iterations of giving feedback, quickly I think Grok got a close facsimile of my style but still content of what it actually wrote was lacking. It still wanted to insert all those freaking em dashes, idk why AI is so obsessed with them, but I kept making it remove them all and tweaking tone. While I write this article (this is not ai generated) I'm also reviewing what it wrote and giving it feedback so hopefully in the future it can do more of the initial heavy lifting.

I say all this because people are quick to use AI out of the box to do creative work for them, but in reality creative AI work needs a body of work to pull from already to get an original voice or visual. This is why so much AI stuff looks so much the same because it does take considerable time and effort to train any AI depending on the complexity of what you want it to understand. For me as a life long artist with a strong identity and years of experience, there are so many soft skills and different views of the world I hold that AI just doesn't understand initially and until I imprint on it my perspective, it will just give me generic outputs. I think with this fact, even with the proliferation of AI, creatives who leverage this to help get the things they don't like doing done like marketing strategy, hashtags for optimized seo, bios to tell people who you are, websites to showcase all your work etc AI can become a complimentary tool for any artist and creators growth, not competitive, even if AI isn't used in their final art.

~~

This leads into my ongoing experiment with what I'm calling Liquid Glitch which is my Midjourney video experiment blending surreal, hyper-saturated visuals with moody ambient sounds all wrapped in short loops that hit like visual ASMR for the scrolling zombies out there. While they are designed to capture the ohhhh shiny brain, they are not just pretty videos. These videos are a culmination of a few weeks of work of me training Midjourney's video feature to animation and output movements it wasn't really obviously trained on before mixing different materials and shapes that generate these uncanny valley vibes that feel like alien dreams invading your feed.

View video works here:

https://zora.co/@empresstrash

https://www.tiktok.com/@empresstrash

Simple breakdown of how I trained it: I prompt Midjourney with basic images. I use said images to create video with prompts of what I would like to see happen in that video. I flag all these with their flagging system like, don't like neutral. I only flag those I like that I feel capture close to what I was asking the video for. If I like something but it doesn't match the prompt of motion I was asking for I mark it neutral. If it doesn't match what I asked for at all I mark don't like. Once this is done enough for images and videos, it saves all the data to your customized profile which then starts bringing your unique vision and vibes with each generation when turned on.

You can hone this down further with moodboards. Moodboards are like training a LoRA for those more tech proficient, or for those not it's like a Pinterest of like images. So you can do things like upload paintings, or all the image generations of a certain style so then you can remove the style from the prompt, focus the prompt on subject and turn the moodboard on to do the heavy lifting of style. All of these things make Midjourney training highly customizable to where people ask me my prompts I'm using, and I point out its about training.

To go even one step further, you can go to the tasks tab in midjourney and train images and video from there from a random pool of content. They even have it the more you train you maybe able to get free fast hours, but also this training helps midjourney video get better while also optimizing your midjourney making it unique to you.

My only issue I have with midjourney video is it's heavy moderation of the feminine form. I find the more curvy and womanly the form is the more likely it won't allow a video to generate even if no nudity and the base image is from Midjourney. This on the surface seems right cause blocking "adult content" to keep the model from generating illegal content, but the reality I see with it is denying adult bodies in training doesn't protect children because it naturally skews the training to make underdeveloped more childlike bodies even in possible situations it shouldn't. I have given Midjourney feedback on this, posted about it publicy, and is a driver right now for a lot of my art. In the world of questionable content and as a survivor of childhood exploitation I just don't think adult bodies being depicted full dressed but maybe provocatively has never been the issue, it's childrens bodies being sexualized that is the issue and will continue to be the issue.

I still love Midjourney, I have been using it since May 22, 2022 v2 (yeah I know so many 2s lol), and all the energy, time and customization I've put into mine I think shows and I feel very invested in as a creative. I think this issue will work out as the model matures and is trained up and the Midjourney community and leadership navigate this complex issue which is why I'm vocal to give feedback at this time. I remember in early iterations of Midjourney you couldn't get a curvy woman to save your life or of different races, and now when it generates women I get all sorts of shapes and sizes, so I've already seen improvement with this on the image side of Midjourney and I'm sure we will work through this issue with video in the same way.

~~

I brought up Liquid Glitch but I tangented on Midjourney training and issues, so back to explain the Liquid Glitch experiment. Think fluid, melting forms, glitchy textures, and a dash of rebellion. Then refine the outputs into seamless loops. Add some custom sounds, and boom, you've got content designed to hook the algorithm while sneaking in some artistic disruption with tightly honed surrealism of objects and humanoid forms.

After I felt basic training was completed for video and posting the training videos to Zora with notes, I decided to reboot my kinda abandoned Tik Tok. I've never been a huge huge Tik Tok person, it felt like another iteration of Insta stuff, but I never hated it either. During this reboot though, I've really learned I think how the algo works for me and how to jump start a tik tok for organic growth. It all kicked off around the start of July, right when I was juggling puppy duty and needing something async to keep my creative fire lit. I started hammering Midjourney video with focused training sessions as mentioned above, portability key since I'd switch from desktop to mobile while chilling with the pup. By July 1, I had Liquid Glitch minted on Zora and posted on Tik Tok.

From there, I escalated things quick. Over the next few days I spam posted a ton of these videos with various hashtags to give the system tons of content and data to test and heavily promoted it while using my AI copilot (Microsoft edge version which I like for these things) to help me navigate how the tik tok algo was responding and what to keep doing. This I think was the crucial piece to help me navigate to grow because I started to see the tik tok algo as a dance partner over an adversary (like I've come to feel on other social media like twitter and insta).I started layering in sounds both from tik tok library and custom with heavy ambient feels thinking of these as healing spells for sad girls but really for anyone also in an existential dread pit.

Around July 4-5. I cross 1k followers and slowed down posting after 34 new posts in the flurry (following fibonnaci sequence to send a signal to the algo I am ready for organic growth) and then set to schedule out content for 2 weeks. This is an important milestone cause 1k followers is when full creator benefits unlock for sure and I personally think you and evidence suggests even if not absolutely concise you are flagged in the system as a creator and your content is viewed differently because post 1k followers posts initially go significantly more views than those right before 1k even though the follower difference was like 50 people and they were posted in the same day. However slowing the posting to a sustainable rate and scheduling it out after such a flurry gives the system time to digest all the information it gained from the flurry, calibrate and know its worth doing that cause this creator has regular content scheduled for testing.

According to my AI Copilot's insights, all content on Tik Tok is tested in tiers and it has to pass each tier before being pushed to the next. Each tier represents a total number of potential viewers it puts the content in front of expanding the pool as it passes each tier. For tier 1, content it has to score a total of 50 engagement points to move on to tier 2 and a wider audience pool. It also told me when an account crosses the 1k threshold the system recalibrates and reassess all content in an extended testing period for 1-3 weeks to gather more data which audiences resonate most with the content. So what all the work I did looks like is a major engagement spike, huge drop and then steady climb out (if successful). Once the steady climb of views happens (mine already has hit that point) at any point if one post finds the right stream and audience, the whole account can go viral, with each new posts adding to the fire like a log making the chances even more likely.

The glitch I initiated in the Tik Tok posting portion is playing with the 1k threshold. Initially spam posting a high amount of high quality (by metrics) content pre-1k, before the system could push to a wider audience because 1k is a threshold/gate, in such a short timespan, it pushed the system into conflict for a few days of wanting to push this new content out to a wider audience but not allowed to because of how the system is designed, therefore I think is what contributed also to the spike in engagement as the system resolved the conflict with the solution of putting my work in front of people enough to get me past 1k, then recalibrate post 1k with all the data. Copilot thinks this will cause a slingshot effect, because it says systems tend to overcompensate when situations like this arise, and what was suppression before becomes fuel for potential explosive growth by crossing the 1k threshold as I did. While I'm still waiting for that explosion in metrics, all other things so far it's helped me map out and predict has been true, so I really feel at this time at any moment it could all explode.

I also have a theory I expressed to my copilot during this that maybe the Tik Tok algo is kind of a gatekeeper of the internet in terms of who is considered creator or regular user. My copilot says this is not true, because algos and AI are siloed, but I don't think it's telling the truth in that. I have had for a long while a sneaking suspicion all the different silos are linked in ways we don't always understand whether through shared data pools, cookies, similar user patterns, or a combination of this with other technological mysteries. I feel this suspicion in this instance has been further validated because a week after crossing the threshold to creator in the Tik Tok system, the X system allowed me to verify my creator status which I have been in a limbo for 3 years after I was erroneously suspended for making a Schrondingers cat joke about poison and it took it as me promoting suicide which is a really bad flag to have on your account. When that happened in early 2022, my growth all but stopped, and even after Elon took over, I still was in a weird limbo where I couldn't sign up for creator benefits. I complained of this a lot to many people over the years.

I would like to also point out all of my growth as an artist in crypto over those years from TED talks to many international exhibitions to my first piece to be in a museum here in CDMX was in spite of a long shadowban. Yes I still have a little resentment and will a while but I'm also very grateful to whoever or whatever resolved that issue.

BUT now, a week after Tik Tok flagged me a creator with that 1k benchmark met, I seem to have been released from that algorithimic limbo I was in on X and got a notification that I should set up my payouts account. I did and also was able to verify in the X system as a creator. I'm going to assume it's like Tik Tok and now over the next few weeks I will see a steady increase in growth I haven't seen before as it reasseses my profile. I will keep people updated on that to see if the data matches my hunches which it has so far.

~~

That's Liquid Glitch in a nutshell a combination of Midjourney training, Zora creator coins use case and social media algorhythm that become just as much part of the art as the shiny visuals itself.

If this glitchy ride resonates and you wanna fuel more chaos, hop over to my Zora profile and snag some creator coin. It's the direct way to support the experiments, keeping the workbench alive for more Midjourney madness and social media conquests. Link's right here: https://zora.co/@empresstrash. Ty for scrolling through the saga. Stay kind 🖤

Find more of Empress Trash @empresstrash on X, Farcaster, Tik Tok and most platforms where art is found.