A i ?

AI as a Medium
(written by a real person)

Complaints about AI art remind me a lot of what people used to say about digital art when it was a new medium.

AI’s big win is removing the gatekeepers of ideas! Letting someone express their own ideas from start to finish, no money, tech skills or permissions needed.

Permission,
Technical Knowledge,
Budget

Advice and tips:

Prompts:
few lose focus
too many prompt words constrict

Describe what you would like to see, not exactly what it is, especially when forcing a perspective. For example. I was trying for days to get a perspective inside a pinball machine, zooming didn’t work, nothing worked until I described the look from inside and didn’t mention pinball machine at all. Mentioning that word made it want to make a whole machine each time and not the portion I was looking for.

Describe the specific visualization in detail & avoid naming larger things when trying to zoom.

Prompts are usually limited in character length but you can tweak it without a noticeable difference; The quick brown fox jumps over the lazy dog. The light is bright. The old man is snoring. can be Quick,fox,jumps,sleepy,dog.bright light.man/old,snoring.

Compiling plots;

Creating a plot with images can seem challenging but if you think of stories in visual ways there are few plot lines that EVERYONE knows. For example, starting in a dawn setting, to noon sun, to evening and night and, if especially long, the am again. The SUBJECT is whatever you choose it to be, if you put a moon in every shot, it becomes a character.

When editing you could go frame > frame > frame but then it begins to look like a slideshow (predictable). Alternatively, you could match the music perfectly, each note and image referenced, which flows better but is then a slave to the melody. To make it stand out in a subconscious way, you have to get creative. Here is a good example;

The medley immediately starts, obviously pairing with an image of a sailboat. The images that relate but are not boats start to leave the transitions, the highs/lows shift out of alignment and length to purposefully add a bit of confusion. Only to return when the next sailboat appears. This gives it another layer comfort and trust in the editing. From then on, I mix it all up and pepper the chorus with the sailboat occasionally to staple it down. There are more complex examples, but this one is clear.

Another important aspect to consider where is the viewer is most likely looking and how will that translate to the next scene. If the focus is on the lower left corner and shifts to the upper right corner on the next scene, that can cause disorientation (which can be useful on purpose) for psychologically, smoother transitions try matching the end to the beginning focal point. I use this technique in the video late night horror;

Often, only a portion of the 4 second clip will be usable, cut out the bad parts and slow down whats left. Anomalies will happen, you can cut them out or embrace them, someday we wont have them so current work will by stylized with a yet unrecognized AI nostalgia.

Every move you make
Every bond you break
Every step you take
I’ll be watching you

The last line purposely doesn’t match.
It is about surveillance and jealousy.
There you make you uncomfortable,
emphasizing the meaning.

Make the clips blend together and use drastic contrasts purposefully. Going from dark to bright, simple to complex or slow to fast on high points or up tempo changes. Try to make it flow, i.e. if you are panning out, pan back in then back out to even the flow. Some clips will need reversing. Introduce dramatic visual changes when a new layer of sound is introduced and you will subconsciously assign that visual with that sound throughout the video, even without seeing it again.

For now, the clips fist frame will be the sharpest and then it starts to fade and blend over time. I have found that removing the first part of the first second can remove the sharpness and blend the work with other frames. Remember its the last frame when you reverse. Sometimes you can use the shaper image part in your work.

What works and what doesn’t?

The ‘sweet’ spot in creating videos for me is where the AI’s best abilities overlap my interests. For example, there have been millions of high resolution and film about cars, so making cars is easy. Where it can get difficult is when you want to do something it hasn’t much reference for. If an analogy wont help, try combining two similar things and try to get an approximation. I have had to abandon more ideas than I have made, many more.

One thing I like to do when I am stuck creatively, is to render as many ideas as I can think of and mix that with any unused clips from previous videos and loosely edit with sound just to see how it looks. The randomness of it can usually spark many new ideas while showing what the AI is most comfortable rendering.

Keep an eye out for unexpected results that could occur from unknown double meanings. For example, I was trying to render a scene and in a mall and it kept showing a street. The reason is because the second definition of mall is “a promenade.” So I changed it to shopping mall and also tried others like retail area etc..

Consistency

One way to show a consistent series of videos or images that have the same feel or style is to come up with a prompt that you like and just change the subject or time of day. Copy and paste your prompts that work to a notepad or somewhere else in case of glitches or to use in the future or to even go back to when an idea isn’t working.

Another way is to use many multiple renders of the same clip. One thing I have found that works really well is to render a clip, then later in the same video show a different render of the same clip flipped (if possible) against another of the same, adding a demotion to a recently familiar scene.

Special prompt words;

  • Words toward the front are the most emphasized.
  • Use words multiple times in the same description.
  • Change words if it is being taken to literally or incorrectly.
  • If you want to create a series use a vague descriptor.

Worn or Worn textures This is an easy way to get your work somewhere between sterile and dirty. Adding a layer of realistic wear to your scene or textures depending on it’s placement position in the prompt. *Update, Surface(s) is superior to textures because it still thinks a texture is applied model sometimes (less real). i.e. Worn Surfaces

Animated textures Sometimes this creates amazing moving textures, sometimes it takes the animation word and creates 3d animation style.

Unique Simply adding the word Unique to your descriptions can give you amazing results. Try using it more than once or paired with detailed, imaginary, otherworldly, futuristic etc..

Ominous and Cosmic Horrors (Towards the end, towards the front will likely cause bannable images when taken directly,)

Specific dates Era films used to be the most expensive kind to make. Adding dates to your descriptions can set the theme for the whole video and give it that consistent flavor.

Negative Keywords There are none and, in fact, it has the opposite effect. Once I was like this is perfect but no skulls or -skulls etc. and every time it made MORE. lol. I will update this once that is possible or I am corrected as to a new technique that works.

Colorize or Technicolor Creates interesting stylized effects. Use this to blend other clips together with a similar feel, even if the subject differs.

Monochromatic, Silent Film, Black and White Sometimes the Silent Film one works and sometimes it doesn’t, the problem is that sometimes it makes film reels, literally.

Uncentered Finding that I have to use this at the start as well as Unique Angle, to get the look away from the default centered common results.

Simulation Ok, this is my new favorite. So, when they last upgraded, I realized everything was coming out too realistic, almost everything was a modern living room. I started to say hyper-realustic oil painting to get back to special effects and unrealistic events. Well, Simulation or realistic simulation does even better.

Collage This term used near the beginning can create some interesting visual effects, I encourage unique paired with this as well.

other terms: edifice, LED, led wire, led accents, led panel, animated texture, juxtaposed, ambient, lithograph, engraving, depth, layers, fireworks (works surprisingly well) tableau, imaginary, otherworldly.

Waterflows When used with simulation you can start making water flow skywards etc, not easy or common, but possible

Special effects Random and doesn’t always work, but when it does, the Special effects prompts are the highlights of the film.

Use Chat GPT or any AI chat of your choosing and tell it what you are looking to achieve. Ask it to help you craft 4 or 5 descriptions that are within the 320 character limit. Try them, then ask it to refine. AI is good at helping you rephrase or fine tune a description to give you better visuals. Be vigilant in sticking towards your ideas (to maintain creative control) but don’t turn away from any happy accidents.

What is it to you:

To me it is a medium, like charcoal or pencil. It is what you do with the medium that defines its output. But, unlike other mediums, it can help you make decisions. The goal, in my mind, would be to find a happy medium, where it has control to do things in the way you want them with some compromise.

The best analogys I can think of right now is a conductor or a film director. Another might be creating stained glass where you cut the glass and put it together in a uniqie way, but you didn’t necessarily create the glass.

Benefits:

Artists have to cut corners, that is, they have to work on something within a reasonable amount of time and prioritize what gets detailed the most. In the past, the eyes have been the most ‘focus’ (rimshot). Because the eyes where what most people looked at first and for the longest. AI doesn’t need to prioritize in the same way because it can fill in any level of detail in any way instantaneously, there is no restriction that requires focal prioritization.

Working as an Art Director, then a freelancer, I understand what the gatekeepers to an Idea are like, I was one. Someone could literally have a great idea and not afford the means to produce it. A good analogy would be the print publishing markets exclusivity that has been eroded away by people who can now publish anything easily.

Downsides:

Infringement Art builds upon past art. People may see that as infringement when it is more like a continuum. The song House of the Rising Sun is a good example of work that has traveled through other artists minds;

art seems to come from somewhere and go somewhere else


To create the most meaningful artwork, you must incorporate everything you have seen and heard which is what ai does. Trying to remove that is like asking a musician to create a new song but can take no inspiration from any past work that is still under copyright. Creating exactly similar work to copyright or trademark work is still protected under law and direrect infringement can be he pursued legally.

* Copyright law also allows for the use of copyright material when learning. ‘fair use’

Easy to do; not real work Like riding a bike with gears. If you are riding the bike the way it was designed to be ridden, you would be peddling the same force in the same speed, but changing gears depending on the elevation. I feel that the same applies to this type of work. If one aspect is easy, then spend your time working on the other aspects that are not as easy, such as specific prompts and prompt arrangements and design and flow and consistency.Etc. there is always a place to do the work, even if the work shifts.


Losing Creative Control You can go with better ideas that it comes up with but then stray far form any creative input or control as to its future direction (losing your input is losing your ability to direct).

Consistency As of yet, consistency has been a problem, Some newer software is trying to address this but it is a current common problem. One solution for now is to choose something that could have a variety of results. i.e. UFO sighting’s in stead of sighting.

Evaluating Synonyms There are sometimes more carefully crafted terms that could go unnoticed. For example, wanting to create something more realistic, I had been saying ‘realistic textures’ well, that made it ok, but I figured out it was assuming I was texturing something fake. I have since started using worn surfaces instead.

Extending Clips When you have a perfect clip and want to extend it, you can try to render it loner but it tends to disintegrate, like a copy of a copy etc.. You can render two different clips, reverse one and then the last frame and first frame of the next clip will match when they are put together. slowed to 66% usually looks natural and sometimes 33% does.

Dating At this point is is still necessary to elaborate on the era to AI. i.e. If you say ‘create a cereal box from the future’, it will create a cereal box THEMED with future elements. However when you say ‘create a cereal box that happens to be in the future’ you get the latter. I have used this many times, it will always default to depictions that represent an era unless otherwise told that the scene merely takes place in that era.

Motivation

I have met many CEO’s when I worked in production and Hollywood producers / executives through the Rotary Clubs in Beverly hills/Bel-Air. I have to tell you, some of the people I served in the Air Force with and some of the people I drove Amazon trucks with, are much MORE creative and MORE able to produce work with meaning. They have just been without the means, until now.

You are the only one with your perspective. AI art has become accessible enough that we can start creating meaningful content for one another without the need for big studios or technical barriers. I’ve worked as a Production Assistant and, more recently, as an Amazon Courier. I made a video tribute to each role. What personal jobs or insights can inspire you to share? They say a good folk song consists of three chords and a truth—let’s do more of that!

Let’s see what you can do!

No experience : Starting from nothing you can make videos like mine. Try not to plan it out ahead and let it flow while creating. Render video clips ( make more than you need! It’s better to decide between to good clips than need one more 😉 ), download them, put them in order in a video editor, cut out the bad parts, slow down the epic parts, add transitions and a music track, (Adobe stock has many, ccmixter, bensounds, also, some artists put their original music on youtube for other’s to use.) Render and upload to youtube or wherever. Then start another! : )

Experienced : what r u doing here 😛 (get to work)