AI Video School

Trip with AI: Mind-Bending Psychedelic Experience Generated by AI

—

by

Mike
in

On my YouTube channel, one of the most frequent comments people make when seeing generative AI for the first time is that it looks like being on drugs. I recently finished reading the book, “How to change your mind” where Michael Pollan’s reporting on mind-expansion drugs included taking different mind-altering drugs.

This inspired me to use generative AI to take me on a mind-altering journey.

Tripping on AI involved 4 AI tools:

  1. ChatGPT to give me the visual language of an AI acid trip
  2. Midjourney to imagine reference image points to guide this psychedelic journey
  3. Stable Diffusion Deforum to bring those image references to life as generative AI video
  4. Eleven Labs IO to create a voiceover based on the ChatGPTs prompts

I was going to add AI generated music but I’m not totally sold on it yet and the licensing for YouTube isn’t clear to me yet.

I had a technical goal, which was to time a Stable Diffusion animation to a voiceover. Most of the Stable Diffusion videos I’ve seen go viral aren’t synched to narration (probably because it’s kind of hard to do as I found out) but I felt like I could do it by timing guided images. That meant I would need ChatGPT to write prompts to generate images in Midjourney and a script to corresponded to those images. Then I would generate a timed animation in Stable Diffusion.

Step 1: Ask ChatGPT to imagine and describe an AI acid trip

The first thing I did was ask ChatGPT to imagine what an acid trip would be like. My goal was to have as much of this invented by AI as possible, with little guidance from me. ChatGPT would write the prompts for Midjourney, Midjourney would be used as reference images for Stable Diffusion,

I’d like to create a generative AI version of what an acid trip would look like. I don’t want to use drugs myself but I want to draw on the documented experiences of those who have to create text prompts for Midjourney which I will then use as reference frames for Stable Diffusion Deforum. Can you help me imagine what an acid trip would look like?

Mike

Certainly! I can help you imagine what an acid trip might look like based on documented experiences. However, please note that the effects of hallucinogenic substances like LSD can vary greatly from person to person, and the experience can be highly subjective. I’ll provide you with some general aspects and themes often associated with acid trips, but keep in mind that individual experiences may differ.

ChatGPT

After a little back and forth with ChatGPT (the initial descriptions seemed a little bland so I had to ask for something truly mind-bending) I got the prompts I would use in Midjourney.

PromptAspect Ratio
Dive into a swirling vortex of kaleidoscopic patterns and colors, as reality twists and folds in on itself. –ar 16:916:9
Picture yourself floating through a fractal tunnel, where each turn reveals new dimensions and infinite complexity. –ar 16:916:9
Imagine a surreal landscape with floating islands, upside-down trees, and waterfalls that flow upwards, defying gravity. –ar 16:916:9
Visualize a morphing, pulsating geometric entity, constantly shifting and rearranging its intricate patterns with mesmerizing precision. –ar 16:916:9
Envision a room filled with shimmering mirrors that reflect not just your own image, but alternate versions of yourself from different realities. –ar 16:916:9
Picture a cosmic dance of glowing orbs and celestial beings, swirling in an ethereal symphony of light and sound. –ar 16:916:9

Step 2: Midjourney generates the images

I got a new set of prompts but I also decided to ask for artistic styles to include in my midjourney prompts. The four that ChatGPT suggested were: Surrealism, Op Art, Psychedelic, Abstract Expressionism. 

Here are the initial images I chose for each prompt.

Dive into a swirling vortex of kaleidoscopic patterns and colors, as reality twists and folds in on itself.
Picture yourself floating through a fractal tunnel, where each turn reveals new dimensions and infinite complexity.
Imagine a surreal landscape with floating islands, upside-down trees, and waterfalls that flow upwards, defying gravity.
Visualize a morphing, pulsating geometric entity, constantly shifting and rearranging its intricate patterns with mesmerizing precision.
Envision a room filled with shimmering mirrors that reflect not just your own image, but alternate versions of yourself from different realities.
Picture a cosmic dance of glowing orbs and celestial beings, swirling in an ethereal symphony of light and sound.

Step 3: Eleven Labs IO for the AI voice over

In order to be able to time these guided images to the Stable Diffusion video, I needed to know how long each section of the voiceover would be. So I went to Eleven Labs and created a new voice generated by AI for free. After listening to a few (without using up too much of my free credits) I didn’t feel like an American accent works well for this and I didn’t want a young voice. The one that sounded best to me was a middle-age woman with an African accent. I’m not sure which country in Africa her accent represents, but I like how she sounds and her cadence is good. After listening to the whole thing, I realized the script needed something to wrap it up, so I went back to ChatGPT and asked for 5 options for a concluding line. I liked the last one the most, so I used that to create an image in Midjourney.

  1. Embrace the unknown and allow your senses to unravel as you journey through a kaleidoscope of infinite possibilities.
  2. Lose yourself in this mesmerizing realm, where the boundaries of perception dissolve, and the extraordinary becomes your new reality.
  3. Step into this extraordinary tapestry of imagination and witness the wonders that lie beyond the realm of ordinary existence.
  4. Let the mesmerizing dance of light and form transport you to a realm where reality and fantasy intertwine, leaving you breathless with awe.
  5. As you emerge from this vivid dreamscape, carry the echoes of this extraordinary experience with you, forever altered by the beauty and wonder you have witnessed.
As you emerge from this vivid dreamscape, carry the echoes of this extraordinary experience with you, forever altered by the beauty and wonder you have witnessed.

I generated voiceover files for each section of the video then took note of the durations, which I would need to convert into frames in Stable Diffusion. I’m using 15 frames per second, so I just need to multiply the number of seconds for each clip by 15. I ended up having ChatGPT write a table with the seconds in one column and the frame number in the other column as a handy guide to help match up the guided images to the voiceover.

Step 4. Animate in Stable Diffusion Deforum

When it comes to generative AI, my mantra is “Generate and Iterate.” There was a lot of trial and error.

This was the timing of my prompts, all of which came from ChatGPT or using the /describe function in Midjourney based on an image generated with a ChatGPT prompt. (ChatGPT isn’t super great at prompting image generators.)

{
    "0": "psychedelic mushroom painting, in the style of detailed marine views, psychedelic artwork, atmosphere landscapes, dreamlike illustration, reefwave, vibrant color palette, detailed compositions",
    "105": "colorful abstract spiral artwork created in the style of hypnotic sprites, luminous 3d objects, mastery of perspective, mixes realistic and fantastical elements, colorful fantasy, high resolution",
    "210": "an island hovering into the clouds, fantasy house of trees, floating in the sky, fantasy, mansion, castle, psychedelic artwork, dense composition, cloudcore, vibrant, lively, cloudpunk, richly layered, vibrant fantasy landscapes, urban fairy tale, highly detailed foliage",
    "315": "pulsating geometric entity, constantly rearranging, intricate patterns, mesmerizing precision, abstract graphical design in different colors and styles, futuristic digital art, aztec art, varying perspectives, focus stacking",
    "450": "shimmering mirrors reflect not just your own image but alternate versions of yourself from different realities, mirror rooms, futuristic psychedelia, an abstract painting of people standing in a room with mirrors, hall of mirrors",
    "555": "a cosmic dance of glowing orbs and celestial beings, a night with glowing lights and colorful lights in the forest, science fiction, space, ray tracing, space, galaxy, ethereal cloudscapes, otherworldly illustrations, dreamlike",
    "645": "illustration depicting a small landscape of forest and mountains, fantastical otherworldly visions, vivid dreamscape, echoes of extraordinary experience, beauty and wonder, mind-bending murals, mushroomcore, cloudpunk, 32k uhd, airbrush art"
}

My guided images looked like this:

{
    "0": "/Users/myusername/Desktop/Trippy/gpt3/1a.png",
    "35": "/Users/myusername/Desktop/Trippy/gpt3/1b.png",
    "70": "/Users/myusername/Desktop/Trippy/gpt3/1c.png",
    "105": "/Users/myusername/Desktop/Trippy/gpt3/2a.png",
    "140": "/Users/myusername/Desktop/Trippy/gpt3/2b.png",
    "175": "/Users/myusername/Desktop/Trippy/gpt3/2c.png",
    "210": "/Users/myusername/Desktop/Trippy/gpt3/3a.png",
    "245": "/Users/myusername/Desktop/Trippy/gpt3/3b.png",
    "280": "/Users/myusername/Desktop/Trippy/gpt3/3c.png",
    "315": "/Users/myusername/Desktop/Trippy/gpt3/4a.png",
    "350": "/Users/myusername/Desktop/Trippy/gpt3/4b.png",
    "385": "/Users/myusername/Desktop/Trippy/gpt3/4c.png",
    "450": "/Users/myusername/Desktop/Trippy/gpt3/5a.png",
    "485": "/Users/myusername/Desktop/Trippy/gpt3/5b.png",
    "520": "/Users/myusername/Desktop/Trippy/gpt3/5c.png",
    "555": "/Users/myusername/Desktop/Trippy/gpt3/6a.png",
    "585": "/Users/myusername/Desktop/Trippy/gpt3/6b.png",
    "615": "/Users/myusername/Desktop/Trippy/gpt3/6c.png",
    "645": "/Users/myusername/Desktop/Trippy/gpt3/7a.png",
    "685": "/Users/myusername/Desktop/Trippy/gpt3/7b.png",
    "725": "/Users/myusername/Desktop/Trippy/gpt3/7c.png"
  }

After a lot of iterations and setting changes, I finally had the video below….