And so I’m back here. Social media has been a wash when it comes to A.I. as the corporations freak out over potential legal liability, which means loss of money. Which is the only threat these corporate bastards respond to. So, will be sharing all my A.I. creations here from now on. Forget about social media.

Anyways, here are a couple of micro-films as proof of concept for an A.I. companion horror story that I am developing.

Let’s be honest about something most music fans aren’t ready to hear: the mainstream music industry didn’t get blindsided by generative AI. It spent the last twenty years rolling out the red carpet for it.

There is a massive amount of panic in creative circles right now about platforms like Suno or Udio “killing the soul” of songwriting. But you cannot automate the soul out of a medium that had already systematically purged it in the name of algorithmic optimization.

Mainstream pop didn’t fall to technology. It formatted itself for execution.

The Grid Built the Runway

For two decades, commercial pop music has traded organic human friction for digital perfection. We replaced the natural, breathing micro-imperfections of a live rhythm section with rigid quantization. Every kick drum, snare hit, and synth line in a modern hit lands precisely on a mathematically perfect grid.

We traded dynamic range—the beautiful contrast between a quiet verse and an explosive chorus—for brickwall limiters that squash audio until it’s a flat, shiny, loud-as-possible block of sound engineered to cut through smartphone speakers.

When you strip out acoustic space, complex harmonic progressions, and structural unpredictability in favor of loop-based timbres and textures, you create the perfect training data. The industry turned pop into a math problem. Why are we surprised that a computer figured out how to solve it?

The Paradigm Shift: From Builder to Director

As an AI creator and composer, I’m not interested in the hand-wringing. The landscape has shifted, and the future belongs to those who understand how to use the tools rather than those mourning a lost paradigm.

When I take my own human-authored intellectual property—my lyrics, my chord structures, my melodic ideas—and use Suno to re-skin them into a flawless 1970s analogue mix or a pristine 1980s synth-pop aesthetic, the creative dynamic changes.

The AI isn’t the artist. It is a hyper-efficient session band and a state-of-the-art vintage recording studio wrapped into a single prompt box.

The workflow shifts from the granular mechanics of tracking instruments to the role of a Creative Director or Executive Producer. You hold the foundational DNA of the song—the emotional intent and the narrative—and you use the technology to dictate the exact historical sonic coat of paint that best serves the vision. It democratizes the studio, leaving the strength of the original song as the only thing that actually matters.

The Shadowy Museum Loop

Will there be a generational backlash? Absolutely. When the digital grid becomes entirely saturated with fully synthetic, automated content, human imperfection will become a premium luxury commodity.

But don’t expect a brand-new creative renaissance. Because the current industry infrastructure doesn’t cultivate or fund raw human ensembles, younger listeners looking for grit and melody are digging backward, not forward. They are reaching for the sounds of the 1960s through the 1990s.

The tragic reality is that we aren’t heading toward an era of radical new musical movements. We are heading toward a permanent, hyper-polished museum loop—a landscape where the past is permanently present, and new art is often just a beautifully curated simulation of a once glorious golden age.

Mainstream digital pop engineered its own predictability, and AI is simply filling the void. You can either spend your energy panic-legislating the inevitable, or you can grab the tools, bypass the existential crisis, and get to work re-skinning the future.

Follow me on SUNO.

Standard AI art is boring. This is why we rejected Dada, embraced the uncanny, and synthesized the Bio-Mechanical Chrysalis Void.

Part 1: We Tried to Generate a Cockroach. It Ended Here.

We started this exercise with a seemingly simple instruction: create a high-fidelity reference image of an anthropomorphic roach.

The response from standard systems was not “uniqueness.” It was a failure of generation. It was an ontological malfunction. Standard moderation systems, the police force of generic AI content, flagged a suit-wearing insect as an unacceptable ‘edge’ case.

Cynical Fact: The most boring thing about ‘art’ in 2026 is its safety. Dada is a century old, but AI systems still can’t handle a creative instruction that pushes outside the smoothed, family-friendly ‘slop’ that defines the current moment.

So, we stopped asking permission. We leaned into the malfunction. The Socratic dialogue we launched—rejecting historical labels and analyzing the current visual conditions—led us directly to this single, inevitable, and disturbing image.

Part 2: Beyond ‘A+B’—Defining the New Synthesis.

Our dialogue identified three arenas of innovative visuals: Microscopic Textures, Unstable Glitches, and Post-Human Symbiosis. You righty recognized that ‘Synthesis’ is the most powerful.

Standard ‘Conceptual Art’ in 2026 is just the same Dada concept re-heated. What is truly cutting-edge is not just “human + machine.” That’s just cyberpunk cosplay.

The Aha! Moment: The New Synthesis is about creating a Symbiotic Subjectivity. It’s about a new form of life or architecture that is neither purely organic nor purely synthetic. The uniqueness is not just “A+B.” It is the creation of a new, single, unified, and unsettling ecosystem.

Part 3: The KMAT 2026 Aesthetic: The Symbio-Cranium_01

We present the header image not as a simple picture, but as the first physical manifesto of this new symbiosis.

VISUAL BREAKDOWN:

• Scaffolding: An ancient, petrified human skull. It is bio-archaeological, not sculpted.

• The Catalyst: Living, vibrant blue-green bioluminescent fungi and delicate micro-circuitry traces. They are not on the bone; they are integrated into its neural pathways, acting as a new neural network.

• The Soul (Micro-Architecture): Inside the empty eye socket, we do not see standard digital glitz. We see intricate micro-gears, turning—they are powered by and driving the symbiotic interaction.

• The Void: There is no stand. No plaque. No ‘museum reflection’. Just an intense, microscopic focus on the integration point.

This is not a depiction of future-primitivism. This is a cold, highly specified, visceral, and beautiful documentation of a new, unified, post-human existence.

Part 4: Conclusion: Embrace the Unsettling.

This entire exercise, which began because we couldn’t generate a reference image of a simple roach, proved that standard creativity is dead.

Standard AI art is smoothed slop, designed to make you feel safe. It is Dada for babies.

The KMAT 2026 Aesthetic Manifesto is about celebrating the ontological fracture. It’s about the integration of Microscopic Decay and Microscopic Glitch into a single, functional Symbiosis. The future is cold, synthetic, and symbiotic. If that makes you uncomfortable, then you are paying attention. Meaning is only found on the edge of the fracture.

The internet is currently vibrating over Seedance 2.0. You’ve seen the clips: photorealistic combat, liquid-smooth physics, and A-list likenesses doing things that usually require a $200 million budget and three years of post-production. The viral narrative claims these were birthed from “two-line prompts.”

As someone who spends more time in the AI trenches than is probably healthy, I’m calling “bullshit” on the two-line prompt. These aren’t magic spells; they are the result of sophisticated multimodal stacking—reference videos, image-to-video anchoring, and iterative “skinning.”

But the technical lie isn’t the story. The real story is the total devaluation of the spectacle.

The Race to the Bottom

If a teenager in a bedroom can generate a sequence that rivals Transformers for the price of a mid-tier subscription, then “Big Robots Hitting Each Other” is no longer a commodity. It’s CGI Slop.

We are entering an era of Spectacle Inflation. When the cost of a high-end visual effect drops to near-zero, the value of that effect also drops to zero. Hollywood is panicked about copyright, but they should be panicked about obsolescence. Their monopoly on “The Impossible” has evaporated.

The Phone vs. The Cinema

There is, of course, a fidelity gap. On a five-inch smartphone screen, Seedance looks like a miracle. On an IMAX screen, the temporal flickering and latent noise would be nauseating. But that gap is closing.

The real kicker isn’t that AI will make movies better; it’s that it will make them cheaper and noisier. Studios will use the “savings” from reduced CGI costs to jam five hundred more explosions into a two-hour runtime, unaware that the public is already hitting “peak sensory overload.”

The Return of Human Friction

When the visual flex is free, what is left to buy?

1. Intent: The “why” behind the shot, not just the “how.”

2. Performance: The soul that a diffusion model can mimic but never inhabit.

3. Human Friction: The knowledge that someone actually risked something to make the art.

The irony of Seedance 2.0 is that it might actually save cinema by making the “digital spectacle” so common it becomes boring. We are going to see a massive pivot toward narrative cynicism and “real” filmmaking—because when everything is possible, nothing is interesting.

I recently dug up a relic from my personal archives: a lyric sheet for a song titled “I’m All Alone,” dated 1979. Looking at those handwritten lines, I realized I had absolutely no memory of the original melody. The 19-year-old me had a vision that the 65-year-old me couldn’t quite recall.

Rather than trying to “find” the old tune, I decided to write a brand new one. I sat down with my guitar, worked out a melody that felt right for the lyrics today, and recorded a basic acoustic demo.

The Social Experiment

Before jumping into production, I posted a photo of the original lyric sheet on social media and asked a simple question: What should this sound like?

The responses were eclectic, but three specific suggestions stood out: 10cc, Elvis Costello, and ABBA.

The Process: From Acoustic to AI

I used Suno’s upload feature, feeding my acoustic demo into the system to ensure the AI stayed faithful to the melody and structure I had just composed. The goal wasn’t to let the AI “write” the song, but to let it arrange and produce it based on those three distinct stylistic prompts.

The Results: A Trip to the Early ‘70s

• The 10cc Version: Instead of the lush, ethereal layers of their later work, the AI went straight for the “Rubber Bullets” vibe. It’s zany, high-energy, and carries that specific brand of 1973 art-glam that feels both clever and slightly chaotic.

• The Elvis Costello Version: This one captured the New Wave snap perfectly—biting, organ-driven, and full of the melodic urgency you’d expect from This Year’s Model.

• The ABBA Version: This was a pleasant surprise. Rather than the “Sad Disco” of their later years, the AI channeled their early glam style. Think “Waterloo” or “Honey, Honey”—punchy strings, bright piano hooks, and that unmistakable early-70s Eurovision stomp.

Final Thoughts

As a composer and songwriter, I find this incredibly exciting. We are moving past the era of AI as a novelty and into an era where it acts as a virtual session band. I provided the DNA—the lyrics and the melody—and the AI provided the studio and the “period-correct” session players.

The result is three versions of a song that sat silent for nearly 50 years. I’m satisfied, I’m inspired, and I’m ready to see what else is hiding in my old notebooks.

Follow me on SUNO.

The Luddite Loop: The Darkroom of the Algorithm

With “The Solarized Siren” and “The Glass Tear Protocol,” we moved from the collage-logic of Ernst into the high-contrast, avant-garde darkroom of Man Ray. If Ernst was about archeology, Man Ray is about the transformation of the mundane into the mysterious through the lens of the “uncanny”.

The Solarized Siren: The Body as Instrument

Inspired by Le Violon d’Ingres, “The Solarized Siren” was an experiment in maintaining high-contrast noir aesthetics while executing a surrealist pivot. We utilized the Veo model to simulate the Sabattier effect—the metallic, glowing “halo” of solarization—in motion. The “disobedience” here was forcing the AI to treat the human form not as a generic figure, but as a marble sculpture that weeps pearls from violin f-holes, defying the digital tendency toward “slop”.

The Glass Tear Protocol: Mechanical Sorrow

In “The Glass Tear Protocol,” we explored “Object-Substitution,” a core tenant of Ray’s work. By replacing liquid tears with glass marbles that sprout ticking metronomes, we layered clashing visual and rhythmic tempos. The challenge was a technical one: keeping the macro-detail of the eye consistent while the tears transitioned from inanimate glass to complex clockwork.

The Rayograph Feedback

Man Ray often allowed the medium to dictate the art—placing objects directly onto film to create “Rayographs.” In the KMAT loop, we treat the AI’s “happy accidents” as our modern silver halide. We are looking for the moment the pixel breaks, the moment the logic of the render fails, and the “disobedience” of the machine creates something authentically surreal.

The Luddite Loop: Tactile Irrationality

With “The Loplop Protocol” and “The Velvet Insolence,” the KMAT experiment moved away from the cool, displaced logic of Magritte into the gritty, tactile world of Max Ernst. If Magritte is about the mystery of the seen, Ernst is about the archeology of the hidden. To capture this, we had to abandon the slickness of digital video and embrace the “frottage” of the machine.

The Loplop Protocol: Peeling the Sky

Inspired by Ernst’s bird-headed alter ego, “The Loplop Protocol” was an exercise in texture. We pushed the Veo model to move beyond the predictable “AI slop” by introducing structural nightmares: trees adorned with human ears and a sky that peels away like layers of physical parchment. This wasn’t just a visual trick; it was a deliberate attempt to use AI to mimic the physical act of collage—layering incongruous textures until the reality of the medium itself dissolves.

The Velvet Insolence: Architecture as a Threat

In “The Velvet Insolence,” we leaned into the “collage-logic” of Une Semaine de Bonté. By replacing city streets with flowing rivers of red velvet and having ladders lean against nothing but low-hanging clouds, we forced the AI to ignore the basic laws of physics. The goal was to create a sense of “insolence”—a structural defiance where the mundane city is treated as a subconscious tapestry that can be unraveled at any moment.

The Frottage of the Pixel

Just as Ernst used rubbings to reveal the hidden “geography” of a surface, we used these experiments to reveal the hidden “geography” of the algorithm. We are looking for the grain, the glitch, and the tactile error that proves the machine is capable of more than just imitation—it is capable of a beautiful, violent irrationality.

The Luddite Loop: Beyond the Literal Prompt

In the early stages of The Luddite Loop, the goal was to prove that generative AI is more than a high-speed Xerox machine for the subconscious. To exploit the true potential of AI art, one must move past the literal and into the conceptual. Our first three videos—“Immortal Cabinets,” “False Reflections,” and our Magritte-influenced “Train” sequence—serve as a manifesto for this approach.

Magritte’s Shadow: Conceptual Displacement

Magritte’s influence on KMAT is never about literal reproduction; it is about the logic of displacement and the confrontation of incongruous subjects. In the “Train” sequence, the aim was to capture a specific “clash”—the arrival of a heavy, industrial force into a domestic, quiet space. The challenge with probabilistic generative AI is its tendency to “smooth over” these contradictions. We had to fight the model to maintain the rigid reality of the objects while ensuring their placement remained fundamentally “wrong.”

The Erasure in “False Reflections”

“False Reflections” provided a masterclass in the value of the “happy accident”. While the initial concept was a symmetrical swap between a man and his reflection, the AI delivered a predatory erasure. The original character didn’t just trade places; he vanished into the silver void. Rather than correcting this “error,” we embraced it. It transformed a simple surrealist trope into a piece of body horror that felt more authentic to the medium’s inherent instability.

“Immortal Cabinets” and Incongruity

In “Immortal Cabinets,” we leaned into the core of Magritte’s philosophy: the juxtaposition of two subjects that simply do not belong together. The difficulty here lies in source fidelity—getting the AI to treat the cabinet as a consistent, mundane physical entity while its interior life defies the laws of space and time. It is about maintaining that “Magrittean” tension where the impossible is rendered with the same flat, undeniable conviction as the everyday.

Embracing the Probability

As a “restless creative,” I treat the probabilistic nature of AI not as a bug to be fixed, but as a feature to be exploited. Every glitch and “misinterpretation” is a potential “happy accident” that pushes the work closer to a true surrealist vision. We are looking for the mystery within the machine, and we are letting it speak.

If you’ve been following my experiments at KMATSG, you know I’m a creative first and a technician second. But lately, being a creative means also being a part-time IP scout. My latest track, “Deep Down,” just went live, and while the music came together beautifully, the visual journey was a comedy of technological errors.

Here is the breakdown of the AI-augmented workflow that brought this 90s-tinged dream pop track to life.

1. Excavating the Past (Gemini)

The lyrics for “Deep Down” weren’t typed into a prompt; they were rescued. I used Gemini to extract text from a handwritten note I’d tucked away in the 1980s. Seeing 40-year-old ink converted into digital data was the first step in bridging the gap between my analog history and my digital present.

2. The Foundation (iPhone)

Technology is only as good as the raw soul you feed it. I recorded a simple acoustic demo on my iPhone to lock in the melody and the emotional core of the song before handing it over to the algorithms.

3. The Sound (Suno)

Using the demo and the 80s lyrics, I utilized Suno to generate a lush, 90s-inspired dream pop/shoegaze arrangement. It captured that specific wall-of-sound resonance I was looking for—drenched in reverb and nostalgia.

4. The Visual Wall (Veo Challenges)

This is where things got “ridiculous.” I wanted a powerful, atmospheric video featuring my AI avatar Enya in an arctic, aurora-lit setting to match the ethereal tone of the music.

However, Veo flat-out refused. Because of the ongoing legal standoff between Google and Disney, the combination of “platinum hair + blue dress + snow” triggered an automatic block. The AI decided Enya’s look was too close to a certain Arendelle princess for comfort.

The Workaround: To bypass the “third-party interest” filters, I had to pivot the aesthetic:

• The Palette: Shifted from ice-blue to deep emerald greens and iridescent purples.

• The Vibe: Leaned away from “Fairytale” and toward “Cosmic Shoegaze.”

• The Result: A 30-second atmospheric loop that actually fits the moody, “Deep Down” vibe much better than the original concept would have.

5. Deployment

The final product—a marriage of 80s lyrics, 90s sound, and 2026 tech—is now live. You can find the full experience on Suno, Instagram, and TikTok (@kmatsg).

The “perfect” AI track doesn’t just happen in a vacuum—it’s usually the result of a mid-Atlantic wrestling match between human intent and algorithmic stubbornness. My latest hook, “At the Club,” is a prime example of how the creative process has shifted from “linear production” to a “recursive loop.”

Here is the breakdown of how a one-minute disco-funk experiment became a fully realized audiovisual asset.

Phase 1: The Sonic Foundation (Suno)

It started with a raw audio demo titled “Disco SUX!”—a skeptical starting point that I fed into Suno to see if it could find the groove.

• The Prompt: I pushed for a classic 70s disco-funk aesthetic.

• The Result: Suno spat back a high-energy, authentic-sounding track that ran just over a minute. The “SUX” irony was lost; the funk was undeniable.

Phase 2: The “Uncanny Valley” Roadblock

The goal was to give the track a face. I moved to Gemini to generate visuals using the Nana Banana model.

• The Problem: The initial results were a disaster. We’re talking “AI plastic”—that overly smooth, hyper-saturated skin texture and those vacant, “uncanny valley” eyes that scream fake.

• The Pivot: Instead of settling for digital mannequins, I stayed within Gemini to brainstorm more nuanced, textural prompts. We moved away from generic “disco imagery” toward specific lighting, grain, and action-oriented descriptions.

Phase 3: Bringing “Kam” to Life (Veo)

Taking those refined prompts, I migrated to Veo. Static images weren’t enough to capture the disco energy; I needed motion.

• The Build: I generated 36 seconds of footage featuring the character “Kam” at the club.

• The Technique: By repeating the core prompts while varying the specific actions (the dance, the lighting shifts, the crowd interaction), I was able to extend the sequence without losing visual consistency.

Phase 4: The Final Merge

The loop closed back where it started. I uploaded the final Veo-generated footage back into Suno to anchor it to the original “At the Club” audio.

The Result: A cohesive, high-fidelity hook video that feels intentional rather than generated.

Check out the Hook over at SUNO.

Beyond the Hype: The High Cost of Killing the Uncanny Valley

The Narrative vs. The Reality If you listen to the AI hype cycles, generating a photo-realistic character in 2026 is as simple as typing a dream into a box. But as we’ve seen with the Year of the Horse series (see images and videos below), the truth is far messier. To get “Kam” to look like a person—and not a plastic mannequin trapped in a digital painting—you don’t just prompt; you troubleshoot.

The “Plastic” Trap The default state of high-end models like Nano Banana Pro is perfection. And perfection is the enemy of realism. During this session, we fought a recurring glitch where the AI defaulted to “Beauty Filter” smoothness. The result? A digital painting that looked “fake as hell.” (see above)

To break the Uncanny Valley, we had to pivot from “high-quality” prompts to “High-Grit” logic:

• Subsurface Scattering: Forcing the light to bleed through the skin rather than sitting on top of it.

• Optical Failure: Using 35mm and 50mm lens profiles to introduce natural distortion and bokeh.

• Intentional Flaws: Specifying humidity-induced flyaways, micro-pores, and mismatched white balance.

The “Two Steps Forward” Glitch We also encountered the technical friction of the tools themselves. From “tool hallucination”—where the AI burns your Veo video quota when you asked for a static image—to the struggle of maintaining Source Fidelity over a long chat thread. Every “step forward” in character consistency often requires “three steps back” into the settings to re-anchor the model’s logic.

The KMAT Takeaway Realism is earned through technical persistence, not automated by the algorithm. If we want AI to be a tool for “restless creatives,” we have to factor in the hidden obstacles. The Year of the Horse isn’t just a theme; it’s a metaphor for the galloping speed of the tech—and how hard we have to pull the reins to keep it grounded in reality.

This short—’The Game’—features my recurring character, Kam. In theory, it’s a high-octane sci-fi thriller. In practice, it’s a showcase of the Uncanny Valley.

The Misinformation Loop:

• The Hype: ‘You can make a movie for the price of a laptop!’

• The Truth: You can make a series of loosely related, slightly-off-looking clips that require massive amounts of post-production to feel even remotely human.

• The Process: We are currently at the ‘8-second wall.’ Attempting a cohesive narrative involves fighting the model at every turn to maintain internal logic.

Stop comparing the cost of a prompt to the cost of a movie. You’re comparing a slot machine to a craft.

#KMAT #TheLudditeLoop #AIFilm #Misinformation #TruthInAI”

We need to move past the cynical “AI will replace film” discourse. Instead of a replacement, we’re witnessing the birth of AI Animation—a distinct genre where the medium’s limitations are part of the message. Why chase the “real” when we can explore the “bizarro”?

The Backstory I recently dug up a notebook from 1981. I was 20 years old, living in the pulse of the New Wave movement, and heavily under the influence of The Police. The song was titled “Only Want You,” and in my head, it was always destined for that crisp, syncopated, “new wave - reggae” fusion that defined the era. For forty-five years, it existed only as ink on lined paper. Today, I used Suno to see if the AI could find the frequency of my 20-year-old self.

The Creative Direction To get the “Sting/Summers/Copeland” DNA, I had to act more as a “creative director” than a technician. AI tends to default to modern, over-produced “bro-reggae” or generic pop. To strip that away, I used a heavy list of Negative Prompts—no auto-tune, no sub-bass, no synths—to force the model back into the analog 80s.

My goal was the “Core Sound”: staccato guitar chords on the off-beat, atmospheric delay, and that high-energy, splashy drumming.

The “Less is More” Breakthrough: Lessons from the Session

The biggest takeaway from this experiment wasn’t about what I added, but what I removed. In the first few passes, I went heavy on the meta-tags, trying to micromanage every beat. The AI responded by going overboard—overwhelming the song with instrumentation that felt cluttered and buried the song’s identity.

I had to play the role of a traditional producer, trimming the instructions down to the bare essentials to give the machine “room to breathe.”

The Human-AI Collaboration in Action:

• Editorial Control: The AI frequently stumbled on the lyrical phrasing. Correcting those mistakes wasn’t just about “fixing errors”; it was about ensuring the rhythmic cadence of my 1981 self remained intact.

• The “Ghost” Hook: Once I cleared the clutter, the machine surprised me. It generated a catchy keyboard hook that I hadn’t planned for. It wasn’t the “reggae trio” sound I originally envisioned, but it fit the 1981 vibe perfectly.

• The Verdict: This wasn’t a “press a button and receive art” moment. It was a genuine collaboration. I provided the history, the structure, and the guardrails; the machine provided the “session flair” and the execution.

Even with the most powerful generative tech, the most important tool in the studio is still the human ear.

SUNO link

They sell us the dream in 15-second trailers. You’ve seen them. “Text-to-Movie!” they shout. “Just type what you see in your head, and the AI will manifest it in 4K resolution!”

It’s a seductive promise. Especially for a creative like me. I’ve spent 30+ years scoring films and writing stories. The idea of bypassing the budget meetings, the casting calls, and the logistics to just create? That’s not just a tool; that’s a superpower.

So, I tried to use it.

The Vision My concept was simple. I have a character, “Po.” I wanted a cinematic close-up of her performing a K-Pop routine—high energy, peace signs, the works—and then speaking directly to the camera with a Singapore Chinese accent.

I had the image. I had the prompt. I hit “Generate.”

The Wall Reality hit me in the face, not with a rendered video, but with a “Safety Policy Violation.”

Apparently, asking for a “K-Pop routine” triggers the AI’s morality police. The algorithms are so terrified of “suggestive movement” (or deepfakes) that they simply block the concept of dancing women entirely. It didn’t matter that my intent was wholesome fun. The machine said No.

So, I negotiated. I stripped away the dance. I removed the specific movements. I asked for a pose.

The Spin Then came the second horseman of the AI Apocalypse: The Spinning Wheel of Death.

I watched a loading circle spin for 20 minutes. I refreshed. I waited. Nothing. The server had choked on the complexity of my request. I was burning credits, burning time, and—crucially—losing Momentum.

As a storyteller, I believe in the Necessity of Momentum. Creativity is a flow state. When you have to stop every 3 minutes to troubleshoot a server error or rephrase a prompt to appease a safety filter, that flow dies. You stop being a Director and start being an IT Support Guy.

The Pivot I am not a technician. I don’t have a $5,000 rig to run open-source models in my basement. I am a Creative.

So, I did what Creatives always do: I cheated.

I realized that if I couldn’t get the “One-Click Magic,” I had to change my approach. I stopped trying to make the AI “act” and started treating it like a chaotic stock footage library.

• It couldn’t handle the dance? Fine. I generated a static “talking head.”

• It couldn’t sell the energy? Fine. I’ll use my music to do that.

See the final video below.

The Reality Check Is the “AI Short Film” a pipe dream? Yes and No.

If you think you’re going to type a prompt and get The Matrix, you are dreaming. That is pure hype. But if you are willing to get your hands dirty—to stitch together broken scraps, to hide the glitches with clever editing, and to let the sound carry the weight—then it is very real.

The AI didn’t make the film. It just gave me the raw materials. The “Filmmaking” still happened in the edit, in the score, and in the choices I made when the machine failed.

The tech isn’t ready to replace us. It’s barely ready to help us. But for now, I’ll take the scraps.

Subscribe now

In the realm of speculative fiction, we often discuss the “Singularity” as a distant horizon. However, in the niche of generative synthesis, the horizon has arrived in the form of Nano Banana Pro. While the name carries a deceptive, almost corporate-whimsical lightness, the underlying architecture is a brutalist powerhouse of image composition.

For those of us accustomed to the slow burn of world-building, this tool represents a radical compression of the “concept art” phase. It is not merely an illustrator; it is a high-fidelity translator of intent.

1. Architectural Capabilities: Text-to-Visceral

Nano Banana Pro operates on a “Nano Banana” foundation, specializing in high-density text rendering and style transfer. Unlike its predecessors, which often struggled with the physics of light in complex sci-fi settings, this model understands the specific “grit” required for high-concept visuals. It excels at:

• Iterative Refinement: It allows for conversational “tweaking,” much like a director giving notes to a cinematographer.

• Compositional Multi-Threading: You can feed it multiple reference images to synthesize a specific aesthetic—say, the brutalism of Andor mixed with the bio-organic horror of Scorn.

2. Proof of Concept: Prompts for the Genre Analyst

To test the model’s capacity for atmosphere over mere “prettiness,” consider these prompts:

• The Horror/Sci-Fi Blend:

  

“A wide-angle shot of a derelict generation ship’s hydroponics bay. The vegetation has evolved into bioluminescent, predatory vines entwined around rusted cryo-pods. Lighting is high-contrast chiaroscuro, 35mm film grain, evoking a sense of claustrophobic isolation and ancient decay.”

• The Fantasy/Political Thriller:

  

“A high-detail architectural render of a sprawling, vertical slum city built into the side of a crystalline mountain. The upper levels are opalescent and sharp; the lower levels are shrouded in smog and neon-blue chemical flares. Wide shot, hyper-realistic, emphasizing the vertical class divide.”

3. The Inhibitor Chips: Moderation and Limits

No technology exists without its “fail-safes.” In the case of Nano Banana Pro, the internal moderation—while necessary for corporate compliance—acts as a creative ceiling.

• The Political Barrier: The system will not render key political figures. For those writing contemporary political horror or alternate history, this requires creative workarounds (using archetypes rather than specific faces).

• The Safety Filter: While it handles “horror” well, it has hard stops on explicit graphic violence. It favors the uncanny and the atmospheric over the literalism of gore.

4. The Path Toward Version 2.0

To truly serve the needs of genre analysts and creators, the model needs to move beyond “generating” and toward “understanding.”

• Spatial Consistency: The biggest hurdle remains maintaining the exact same character or environment across different angles. We need a “persistent world” toggle.

• Logical Physics: While the textures are perfect, the model occasionally fails the “Escher test”—creating stairs that lead nowhere or limbs that defy anatomy in high-action shots.

Nano Banana Pro is a formidable tool for the belt, provided you know how to navigate its constraints. It isn’t a replacement for the human eye, but it’s a hell of a lens.

As a bonus, here are videos generated (with Veo) using the scifi-horror motif background above as the starting point:

Subscribe now

By K.MAT

If you listen to the current discourse, we are standing on the precipice of the end of art, the end of truth, and the end of the human soul. Again.

It’s exhausting, isn’t it? The certainty that this specific technological leap is the one that finally kills our humanity.

But here’s the thing: we have been “dying” for 2,500 years.

For the past week, I’ve been using the very tool that everyone is currently panicking about (Generative AI) to resurrect the ghosts of panics past. I call this series “The Luddite Loop.” It’s a collection of 8-second screams from history, capturing the exact moment a specific generation decided that the future was “grotesque.”

Here are the exhibits.

Exhibit A: The “Brain Rot” Argument (370 BC)

We start with the OG hater: Socrates. We tend to think of the “screens are rotting our brains” argument as a modern affliction. But in 370 BC, the intellectual elite were terrified of a new, dangerous technology called The Written Word.

Socrates argued that if we wrote things down, we would cease to use our memories. We would become “hearers of many things and learners of nothing.” The Verdict: He was technically right. I don’t remember my best friend’s phone number. But I also don’t have to memorize the entire Iliad to know what happens to Hector. Fair trade.

Exhibit B: The “Soulless” Argument (1811)

Fast forward to the Industrial Revolution. The Weavers of England didn’t just hate the Power Loom because it threatened their jobs (a valid economic fear); they hated it because they believed a machine could not imbue cloth with “spirit.”

In my video reconstruction, our Weaver screams about “Soulless Trousers.” The Verdict: Looking at the current state of fast fashion... he might have had a point.

Exhibit C: The “Mystery” Argument (1927)

When The Jazz Singer premiered, it wasn’t universally hailed as progress. For the stars of the Silent Era, “Talkies” were an aesthetic crime.

My Silent Film Star character captures the specific narcissism of this panic: “If they hear my voice, they stop looking at my face.” The Verdict: We didn’t lose the “mystery.” We just traded pantomime for dialogue. And now, 100 years later, we are using AI to generate silent videos again. Time is a flat circle.

Exhibit D: The “Sweat” Argument (1930)

The American Federation of Musicians waged a literal war against recorded music. They called it “Canned Music“ and argued that if you couldn’t smell the musicians sweating in the pit, it wasn’t art. It was a “Robot’s Lullaby.”

The Verdict: We eventually decided that we liked listening to music in our living rooms more than we liked the smell of frantic violinists.

Exhibit E: The “Lazy” Argument (1975)

Finally, the one that hurts the most. The Math Teacher. “You won’t always have a calculator in your pocket!”

This was the lie an entire generation was raised on. The fear that technology would make us helpless. The Verdict: We do have calculators in our pockets. We also have the entire sum of human knowledge. And yet, we still struggle to calculate the tip at dinner.

The Takeaway

I didn’t make these videos to mock these fears. The economic anxieties were often real. But the cultural panic—the idea that a tool would strip us of our essential humanity—is a repeating script.

We always survive the tool. We just change how we define “human.”

Follow my A.I. explorations on my Instagram or TikTok.
And remember: Nothing is real

Subscribe now

Can a machine feel the humidity? A deep dive into generating “Life in the Lion City” with Suno.

They say nostalgia is a glitch in the human operating system—a longing for a file that has been corrupted or deleted. Today, I decided to see if I could force that glitch into an AI.

I wanted to generate a track that sounded like the Singapore I grew up in (or the one I remember, which might be two different things). I didn’t want the polished, tourist-board pop. I wanted the jangle. I wanted the sweat. I wanted the sound of a humid Saturday afternoon in 1994.

I wanted Kevin Mathews.

If you know the local scene, you know the name. The Watchmen. Popland. The Godfather of Singapore Power Pop. He wrote songs that sounded like the Beatles if they’d been born in Toa Payoh instead of Liverpool. He captured that specific, bittersweet ache of living in a city that tears itself down every five years.

So, I fired up Suno Pro to see if I could teach a neural network to sing about “void decks” without sounding like a GPS navigation system.

The Experiment

The goal was simple: Create a power-pop anthem called “Life in the Lion City.”

Most people use AI music generators wrong. They type “cool rock song” and get garbage. To get the soul, you have to be a director. You have to give the AI a vibe and a history.

Here is the exact “recipe” I fed the machine. Note the specific genre tags—I was trying to trigger the AI’s latent training data on 90s indie rock.

The Prompt:

Style: Upbeat Power Pop, Indie Pop, Jangle Pop, 90s Singapore Indie, Catchy, Melodic, Witty, Beatlesque, Acoustic and Electric Guitar blend, Male Vocals.

The Lyrics:

I wrote these myself. AI is good at structure, but it’s terrible at local details. An LLM doesn’t know the specific rhythm of an Uncle washing the floor at 7 AM. It doesn’t know the “complicated pity” of queuing for food.

I fed it lines like:

Uncle is washing the void deck floor / Same old rhythm as the day before

I grab a Kopi-C and I scan the news / Got the humid weather, post-colonial blues

The Result

The AI spat out a track that... honestly? It scared me a little.

It nailed the Kevin Mathews signature sound. The acoustic guitar intro is bright and percussive—that “strumming while walking fast to the MRT” tempo. The vocals have that slight, nasal indie affectation, not too polished, just earnest enough.

When the chorus hits—“It isn’t always pretty / But it’s the only place I know”—you almost forget it was sung by a server farm in a cold climate somewhere. For three minutes, the machine convinced me it was sweating through a cotton T-shirt on Orchard Road.

Why This Matters

We are entering an era where “Nothing is Real” (you know my motto). But here’s the twist: If the fake song makes you feel a real emotion, does the origin matter?

This track is a simulation of a style from 30 years ago, generated by a ghost in the machine today. It’s a copy of a copy of a memory. And yet, when that slide guitar solo hits in the bridge (very George Harrison, very Cloud Nine era), I tap my foot.

Listen to the Track

The Prompt Recipe for Subscribers:

Want to try this yourself? Here is the exact prompt structure I used. Steal it. Remix it. Make your own fake memories.

• Genre: Power Pop / Jangle Pop / 90s Indie

• Descriptors: Witty, Melodic, Upbeat, Acoustic-Electric Blend

• The Secret Sauce: Use meta-tags in your lyrics like [Melodic Bassline] and [Tempo Slows] to force the AI to arrange the song, not just sing it.

Don’t fall in love with the singer, though. He doesn’t exist.

— K.MAT

“Nothing is real.”

Subscribe now

Three years after I began my A.I. journey, I’m finally at the point where I can give life to some of my creative ideas. At the moment, generative AI video with Gemini’s VEO is coming along decently, albeit with many limitations and flaws. I am exploring the idea of putting together a 10-15 minute short film with the help of VEO and testing out concepts with VEO.

So I had previously written about the concept of me as a time traveller and visualizing myself at significant historical events meeting significant historical figures. This has been occupying my creative time in the last few days - hit or miss, as you might imagine. Watch the video compilation below (about 90 seconds) and let me know what you think.

As you can see, there are still noticeable flaws, character inconsistency being the main one. In addition, I had to work around several prompts that triggered the Gemini moderation gestapo. For example.

The A.I., while perfectly fine with generating this image, refuse to animate children! A pity. This was the same problem with the image below.

Again, the prompt was rejected as a child is depicted. It’s a blanket ban on animating children, which is silly, but there you go.

Time to move on to the next concept in my A.I. journey. Hope you will come along for the ride!

Subscribe now

The music industry just hit a major milestone that many saw as inevitable, yet few expected to happen quite this fast. In a rare display of cross-border unity, North America’s three performance rights giants—ASCAP, BMI, and SOCAN—have officially aligned their policies to allow the registration of “partially AI-generated” musical works.

For years, the industry lived in a state of high-tension anxiety over generative AI. But as major labels transition from litigation to collaboration with platforms like Suno and Udio, the gatekeepers of performance royalties are finally opening the door.

Why This Matters

Until recently, the legal status of AI-assisted music was a gray area. The new consensus is clear: if a human is the “driving creative force” behind a composition, it is eligible for registration and royalties, even if AI tools were used to help realize the vision. This shift mirrors the recent pivot by major record labels who, seeing the success and creative potential of Suno and Udio, are now seeking ways to license their catalogs for training rather than simply fighting the tide.

We are moving past the “AI vs. Human” debate and into an era of AI-Assisted Human Artistry. It’s no longer about whether AI is used, but how it is used as a creative partner.

Realizing “Game”: A New Approach to Composition

To put these new standards to the test, I’ve just released my latest track, “Game.” This song wasn’t just “prompted” into existence; it was the result of a hybrid workflow that reflects exactly the kind of human-led creative process the PROs are now recognizing.

Here is how “Game” came to life:

1. The Spark: It started the old-fashioned way—I woke up with a melody stuck in my head. No algorithms, just a recurring tune that wouldn’t leave.

2. Capturing the Ghost: I hummed those melodies directly into my iPhone voice memos to preserve the rhythm and pitch.

3. Lyrical Sculpting: I used AI to help iterate on lyrics, providing detailed prompts based on the themes of the melody to find the perfect poetic fit.

4. The Realization: I brought my humming, my curated lyrics, and specific style prompts into SUNO to fully realize the track. By acting as the “director,” I was able to steer the AI to match the specific vibe and structure I had hummed into my phone that morning.

The result is a track that feels deeply personal because the core DNA—the melody and the intent—originated in my own mind.

Listen to “Game” on Bandcamp

You can hear the final product of this experiment right now. “Game” represents my exploration into this new frontier where technology acts as the ultimate session musician and co-writer.

As we move forward, the question isn’t whether AI will be part of the charts, but how we, as creators, will use these tools to amplify the songs we’ve always wanted to write.

Subscribe now