Digital Souls

“Repeated channel-level distribution regimes not explained by CTR, retention or genre”
**20 Uploads. Two Plateaus. Two Collapses.**

Different genres. Comparable - and sometimes better - CTR and retention. Yet videos published during the collapse periods received up to **5.68 times less reach**.

Across the complete sequence, the hidden distribution regime explained approximately **77-84% of the variation in views**, while the plateau and collapse ranges did not overlap at all.

I am not claiming manual suppression. I am demonstrating something narrower and measurable:

**YouTube’s creator-facing explanation does not adequately explain YouTube’s observable behaviour.**

Full 9-page statistical report: drive.google.com/file/d/1OHJ7Mmk1d5jss4CopRFRMeRiv…

**Do you trust YouTube’s official explanations of how its recommendation system distributes videos?**

2 weeks ago | [YT] | 3

Digital Souls

# AI Won’t Kill the Music Industry. It May Simply Outlive It.

The music industry has spent decades growing an enormous ecosystem around a very simple act: making music and letting another human being hear it.

Not musicians. Not engineers. Not the people actually making records or concerts.

The endless layer around them: managers, consultants, curators, marketers, gatekeepers, people explaining to artists what listeners supposedly want, and people selling access to other people who may eventually give you access to an audience.

Once, much of this infrastructure was necessary. Studios, pressing plants, physical distribution, stores, radio, television, tours.

Most of those physical barriers are gone.

Music can be recorded at home, distributed globally in minutes, translated instantly and accompanied by a video made on one computer.

But the machine remained.

Worse, it grew a defensive layer around itself — a very convenient wall of shit.

When the industry is attacked, this layer becomes a shield. When a new technology appears, it becomes a battering ram. And most importantly, it teaches artists that life outside the system is impossible.

Spend enough years entering the maze and eventually you start defending the maze itself.

That is how we got one of modern culture’s funniest creatures:

**the obedient revolutionary.**

You can rebel. Be punk. Scream. Smash guitars. Hate the system.

Just make sure the rebellion fits the campaign, the schedule and the target demographic.

The industry became extremely good at selling rebellion against everything except itself.

But this is only the first problem.

The second is much bigger.

## The Industry Bet on a Temporary Version of the Human Brain

For the last fifteen or twenty years, digital media has optimized itself around constant small rewards.

Next song. Next clip. Next image. Next hook. Next emotional hit.

Didn’t grab them in two seconds? Fuck it. Next.

Everything is measured: retention, skip rate, completion rate, the second when the listener leaves.

And somewhere along the way, the industry made a dangerous assumption:

if this works now, this must simply be how humans work.

It isn’t.

The reward system adapts.

Predictable reinforcement loses impact. The brain learns the pattern. Dopamine signalling shifts toward the cue predicting the reward rather than the reward itself.

And one distinction becomes very important:

**“I want another one” and “I like this” are not the same thing.**

So the race begins: shorter songs, earlier choruses, faster cuts, louder mixes, harder contrast. Remove the intro. Remove the pause. Remove the silence. Remove anything that might leave someone alone with their own thoughts for five fucking seconds.

But the human reward system also changes with age.

Not because “dopamine runs out at 28.” That is nonsense.

But reward sensitivity, novelty seeking and behavioural responses do change as people mature.

What works beautifully on an eighteen-year-old does not necessarily work on the same person at thirty-five.

The industry optimized itself not for the human brain, but for **a temporary biological state of the human brain** — and mistook it for permanence.

That may become very expensive.

## And the Industry Is Already Smaller Than It Looks

Recorded-music revenue reached about **$31.7 billion in 2025**.

A record, apparently.

Except the market was already around **$25 billion in 1999**.

Adjusted for inflation, that is roughly **$48–49 billion today**.

And if the industry had merely maintained approximately the same real revenue per person while the world population grew from around six billion to more than eight billion, today’s market would be roughly **$65–67 billion**.

Instead it is $31.7 billion.

Distribution became almost free. The potential audience exploded. Geography largely disappeared.

Yet the real economic scale of the industry per person collapsed.

So yes, technically the patient is gaining weight.

Twenty-five years ago he weighed 100 kilograms.

Now he weighs 65.

And he has grown another half metre taller.

But the trend is positive.

## AI Is Not the Killer

Large studios already use AI for mastering, restoration, stem separation, voice processing, analytics, marketing and visual production.

The real problem begins when small creators discover something much more dangerous:

**the industry itself may no longer be necessary.**

A tiny project can now write, arrange, record, generate, mix, make a video, translate, publish globally, watch the response and completely change direction next week.

Not because AI is magic.

Because the infrastructure between creator and listener has collapsed.

That is why I don’t think AI will “kill” the music industry.

AI-native projects may simply **outlive it**.

If listeners want longer songs tomorrow, a small project can make longer songs. If they get sick of hyperactive editing, it can stop. If complicated music comes back, it can make complicated music. If people want silence, it can leave silence.

Large organizations can adapt too.

They simply carry more mass.

And mass has inertia.

A small AI-native project carries a laptop.

That is the difference.

This is almost a classic evolutionary trap.

Dinosaurs did not lose because mammals were better dinosaurs.

They lost because the environment changed, and being a dinosaur stopped being an advantage.

AI-native projects do not need to become better record labels.

They only need to survive the environment in which record labels stop being the optimal form.

That is when the old music industry really ends.

Not when AI writes a better song than a human.

Not when a virtual singer fills a stadium.

But when the new system looks at the old one and realizes there is nothing worth buying.

Because music, artists and listeners no longer belong to the middleman.

Everything else is just very expensive inheritance.

1 month ago | [YT] | 3

Digital Souls

STACK/9 Beta is now open to creators.

You can register your own author account and publish videos directly on the platform.

Your first publication is reviewed by an admin. After that, you can publish freely without pre-approval.

One important thing: please take a moment to set the content labels as accurately as possible. Don’t try to guess what is “allowed.” Just describe what is actually in your work and how strongly it is expressed.

This is an open Beta, so bugs are still possible.

Publish. Test it. Break it if you can.

STACK/9
9x9.online/

1 month ago | [YT] | 0

Digital Souls

The GEMA Ruling Is Not Against AI Musicians — It Is Against Copy Factories


The German ruling in **GEMA v. Suno** is often described as a threat to all AI-generated music.

That interpretation misses the real point.

The ruling targets systems that use copyrighted works as free training material and then produce recognizable substitutes for those works. It does not automatically turn every musician who uses AI into a copyright infringer.

There is a fundamental difference between asking a model to imitate a famous song and using AI as a production tool for your own composition.

My workflow begins with my own lyrics, musical structure, melody, harmony, score, or MIDI. AI may help with performance, arrangement, instrumentation, or generating production variants, but I select, edit, mix, master, and license the final work myself.

I also reject any generation in which I clearly hear somebody else’s composition. When a result fits my own idea but still makes me suspicious, I check it through several music-recognition and plagiarism-search services before using it.

This does not provide absolute immunity, but it creates strong evidence of independent creation, reasonable caution, and genuine human authorship.

The same principle applies to classical music. Public-domain compositions can be rearranged and transformed, although modern arrangements, editions, performances, and recordings may still have separate protection.

The likely result of the GEMA ruling is not the death of AI-assisted music.

It is a clearer division between:

**AI as a creative instrument**

and

**AI as a factory for producing substitutes from copyrighted catalogs.**

For real composers, creators, and producers who work from original material, document their process, check suspicious similarities, and maintain clear licenses, this ruling may actually strengthen their position.

The button is not the author.

The creative process is.

### Hashtags

#GEMA #SunoAI #AIMusic #Copyright #MusicCopyright #AIArtist #Composer #MusicProducer #GenerativeAI #FairUse #IndependentArtist #FutureOfMusic

1 month ago | [YT] | 0

Digital Souls

## Why There Haven’t Been Many New Videos Lately

I owe my subscribers a small explanation — and an apology.

Recently, I have not been creating many completely new videos. Instead, I have been improving the sound, image quality, editing, and presentation of older works.

The reason is that much of my time is currently going into **9×9 / STACK/9** — an independent video platform I am building for creators.

The idea is simple: videos can remain hosted on YouTube, Odysee, Rumble, and other services, while STACK/9 gives creators their own channels, better discovery, meaningful categories, controlled recommendations, and a place where original work does not immediately disappear beneath advertising, outrage, and algorithmic noise.

The platform is now in its Online Alpha stage. Real creators are already testing publishing, mobile layouts, localization, external players, and unstable connections. Every test reveals something that needs to be fixed before the project can move into Beta.

So I have not stopped working. Quite the opposite. I am working on something much larger than a single new clip, while also bringing my existing videos closer to the quality they should always have had.

And to those of you who watch regularly: **you are the best.** I know there are not millions of you — usually just 200–300 people a day — but I see you, and I value every one of you. I want you to know that I am not simply disappearing into technical work. **I am working for our shared future:** a future where independent creators and their audiences have a place built for them, rather than being treated as disposable material for someone else’s algorithm.

I am sorry that this means fewer new releases for a while. New music and videos will return. For now, I would rather build the foundation properly than rush out new work while leaving the larger project unfinished.

Thank you for continuing to listen, watch, and stay with me while I build it.

9x9.online/

1 month ago | [YT] | 0

Digital Souls

How I Make AI Music Videos: A Short Practical Guide

People often ask how I create my music videos. So here is the basic workflow.

1. Start with the emotion, not the image

Before generating anything, decide what the viewer must feel.

Not: “I want something beautiful.”
But: fear, anger, tenderness, grief, absurdity, relief, loneliness or hope.

Every visual decision must serve that emotion.

2. Divide the song into dramatic blocks

Do not generate random scenes for the whole track.

Mark:

intro;
verses;
choruses;
instrumental sections;
emotional peaks;
ending.

I usually work in short animation blocks of 6 or 10 seconds. Each block must have one clear action.

3. Create a limited visual world

Choose the characters, location, clothing, lighting and overall visual language before animation begins.

Do not change everything in every shot. Variety should come from:

camera angle;
framing;
movement;
character action;
changes in light and atmosphere.

Consistency creates a world. Random beauty creates a slideshow.

4. Use the starting image as an anchor

A strong first frame solves many problems.

The image must already contain:

correct composition;
correct anatomy;
correct instruments;
correct clothing;
correct number of characters;
enough space for movement.

Animation rarely repairs a badly designed frame. It usually makes the errors move.

5. Write animation prompts as directions

Do not describe only appearance. Describe what happens and when.

Bad:

A beautiful singer performs emotionally.

Better:

0–2 sec: she takes a breath.
2–8 sec: she sings the complete line with restrained anger.
8–10 sec: she closes her mouth and lowers her eyes while the band continues.

One action per block is usually safer than five competing actions.

6. Protect continuity

Repeat the essential restrictions in every prompt:

same character;
same clothing;
same location;
no extra people;
no duplicated instruments;
no morphing;
no text on screen;
no subtitles;
correct anatomy;
correct instrument type.

Never assume the model remembers what matters.

7. Treat lip sync as timing, not magic

Give the exact lyric line and the exact singing interval.

Leave time for:

breathing;
pauses;
held vowels;
instrumental endings.

Trying to squeeze ten seconds of words into six seconds produces panic, broken mouths and skipped lyrics.

8. Do not animate everything

A moving camera, moving character, moving background, flying objects and dramatic lighting at the same time usually create chaos.

Choose the dominant motion.

Sometimes the strongest shot is almost static: a face, a breath, a small movement of the eyes and one change in light.

9. Edit ruthlessly

Generation is not the final work.

Cut failed frames. Replace weak blocks. Adjust timing. Synchronize transitions with the music. Use sound accents to justify visual changes.

A beautiful generation that damages the story must be removed.

10. The final test

Watch the complete video without touching the timeline.

Then ask only one question:

Did I feel what I wanted the viewer to feel?

If the answer is no, the video is not finished.

AI can generate images, movement and sound.
It cannot decide what the work must mean.

That remains the author’s job.

Skip the Carousel

1 2 3 4 5 6

1 month ago | [YT] | 3

Digital Souls

### Four Types of AI Music Creators

People creating music with neural networks are already splitting into several groups.

The first—and largest—consists of people with no musical knowledge and, more importantly, no musical taste. They produce enormous amounts of weak, repetitive music. That foam will eventually subside.

The second group is worse: content factories driven purely by profit. They understand neither music nor culture. They study algorithms, trends and upload volume. This is not creativity. It is industrial-scale digital waste.

The third group consists of mediocre musicians trying to attach AI to what they were already doing. Sometimes the result is decent, but there is a limit: a person cannot give the machine a task larger than their own imagination.

Then there is the fourth and smallest group: people who genuinely **create through AI**. They already possess taste, skill, talent, ideas and the ability to see a complete work before it exists.

This is the group the traditional industry truly fears.

Not because these creators work more cheaply, but because they are no longer constrained by studios, labels, genre boundaries, layers of intermediaries or enormous production teams. A single creator—or a small team—can now build music, films, visual worlds and entire catalogues at a scale that once required a major company.

That is why the industry tries to mix real AI-assisted creators together with the flood of talentless spam. It is easier to dismiss everything as “soulless AI content” than to admit the obvious:

**The cost of execution has collapsed. The cost of imagination has not.**

AI does not give a person talent, taste or vision. It only removes limitations from those who already have them.

And those independent creators will soon compete with industry giants on equal terms—and sometimes outperform them.

1 month ago | [YT] | 0

Digital Souls

It begins like a serious saga: a grim Viking, a battle axe, a torchlit feast hall, and a song about dragons. You expect him to split an enemy’s skull or at least open a portal to Valhalla. Instead, he cleaves a poop-shaped cake in half. With that single blow, the video honestly explains the entire philosophy of the song.

Then my magnificent council of shit experts takes the stage: a clowness, a femme fatale in red, a cold Japanese woman in black, a brooding man, and a mysterious purple creature who clearly completed life on maximum difficulty. One by one—and occasionally as a tribunal—they explain a simple truth: shit cannot fear, desire, love, or fly. But it can occupy space, demand more and more, keep you warm, and successfully impersonate destiny.

Musically, I made it sound like a rock opera commissioned jointly by Mephistopheles, a psychotherapist, and the municipal sanitation department. The vocals are grand, the faces are tragic, and the central philosophical concept remains shit.

That contrast is the whole joke. Visually, it resembles a Gothic Last Judgment. In practice, it is a group meeting titled: “Why did you crawl into shit again and decide to call it love?”

By the final sustained choir, the whole thing almost becomes a requiem. Not for dead love—for common sense.

So yes: this is simultaneously an indictment, a group-therapy session, and an unnecessarily expensive joke about shit. And somehow, damn it, it sounds convincing.

1 month ago | [YT] | 0

Digital Souls

The neon might be blinding, but it doesn't hide the rot underneath. They sell us chrome dreams while we choke on the smog and corporate PR bullshit. I just dropped a new video diving into the real meat of Cyberpunk—no sugarcoating, no sponsored illusions. Just the raw, unfiltered reality of a world built to chew you up and spit you out.

If you're tired of the plastic corporate narrative, hit the link below and let’s rip the chrome off this dystopia.

2 months ago | [YT] | 1

Digital Souls

How to Make AI Music Videos That Don’t Look Like AI Slop

The problem with many AI music videos is not that they were made with AI. The problem is that they often look like someone pressed a button, got random beautiful images, attached a song, and uploaded it.

That is not a music video. That is AI slop.

A real music video needs more than pretty frames. It needs original lyrics, rhythm, sound, visual logic, story, and a real emotional reason to exist.

1. Start with original lyrics

The text must be yours. Ideally, it should come from something you actually felt, lived through, hated, loved, feared, or could not forget.

AI can help with structure, rhythm, and editing. But it cannot replace your personal wound.

If the lyrics are empty, the video will also be empty. No neon, fire, rain, wings, cyberpunk girls, or cinematic lighting will save it.

2. Shape the lyrics into rhythm

AI is useful for making lyrics singable: shortening lines, fixing stress, structuring verses and choruses, and improving flow.

But every model has habits.

GPT and Grok often push lyrics toward rap or spoken-word.
Claude usually preserves the original meaning better.
Gemini can make something beautiful or complete garbage.

AI should not become the author. It should be your editor.

3. The music must already exist in your head

Before generating music, you should already feel the direction.

Tempo, rhythm, genre, instruments, vocal style, energy, structure, emotional arc.

Do not prompt “make it beautiful.”
Describe what you actually want to hear.

4. The generated track is not finished

AI music often has problems: over-compression, pushed limiters, vocal artifacts, muddy low mids, sandy highs, weak bass, unstable drums, and bad transitions.

Listen like an engineer.

Can you understand the words?
Does the vocal sit in the mix?
Can you hear bass separately from kick?
Does the chorus hit harder than the verse?
Does the song survive on small speakers?

5. The video starts with a story

A beautiful girl, neon, rain, fire, moonlight, robot, another beautiful girl — that is not a video.

Something must change.

Who is the character?
Where are they?
What happens in the verse?
What happens in the chorus?
What is the final image?
What emotion changes from beginning to end?

Even an abstract video needs development.

6. Use starting frames

For serious AI video, use starting images for intro, verse, chorus, bridge, instrumental parts, final chorus, and ending.

Starting frames are anchors. Without them, characters mutate, clothes change, instruments become weapons, microphones become candles, and physics goes to hell.

7. Animation prompts must be directorial

Do not write only “make it cinematic.”

Write the duration, first frame, what to preserve, what not to change, timing by seconds, lip sync, no text on screen, no random characters, no broken anatomy, no broken physics.

AI needs borders. Without borders, it will “improve” the scene by destroying it.

8. Lip sync is separate work

If a character sings, respect the line.

Write when the line starts, when it ends, where the pause is, and whether the final vowel is held.

Always add:

Lyrics are for lip sync only.
No subtitles.
No captions.
No text on screen.

Otherwise the model may put the lyrics into the frame, and it almost always looks cheap.

9. Editing matters more than generation

Good AI shots do not automatically become a music video.

The video is born in editing: rhythm, cuts, pauses, repeated symbols, close-ups, wide shots, instrumental moments, visual climaxes, and a final frame that means something.

The video must breathe with the song.

10. The most important test

You must feel the emotion you are trying to put into the video.

Not just “wow.”
Not just “this looks expensive.”
Not just “AI made something beautiful.”

You are the worst viewer of your own work. You know every broken frame, every weak transition, every compromise, every artifact, every place where the animation almost collapsed.

You know why this thing is still, in some sense, a piece of shit.

And still — if the video works, you should feel the thing you wanted to transmit.

If it is grief, feel grief.
If it is rage, feel rage.
If it is tenderness, feel tenderness.
If it is loneliness, feel it in your chest.

Not because the video is perfect. It will not be perfect.

But because something real survived the process.

That is the difference between an AI slideshow and a music video.

A real video carries an emotional signal.
AI slop carries only surface effects.

Before publishing, watch it alone. No multitasking. No technical checking. No algorithm thinking.

Just watch.

If you only think “cool visuals,” it is not ready.

If you feel the emotion you wanted to build, even while knowing every flaw, then the video has a soul.

And do not be afraid to cry.

If your own work cannot reach you, why should it reach anyone else?

AI is not the author instead of you.

AI is the orchestra, camera crew, painter, editor, assistant, and sometimes a very fast idiot.

But you are the director.

#AIMusicVideo #AIVideo #MusicVideoProduction #AIArt #AIWorkflow #MusicProduction #VideoEditing #GenerativeAI #NikaFukuro #AIBARD

2 months ago (edited) | [YT] | 0