GENERATIVE AI STUDIO

GENERATIVE AI STUDIO

SYSTEM OVERVIEW =))))

CUSTOM LORAS ARE TRAINED USING CAREFULLY CURATED IMAGE DATASETS, SOURCED DIRECTLY FROM CLIENT OR BRAND ARCHIVES. THESE DATASETS ARE DESIGNED TO STABILIZE SILHOUETTE, MATERIAL BEHAVIOR, AND LIGHTING SO THAT GENERATIVE IMAGERY REMAINS CONSISTENT ACROSS VARIATION RATHER THAN DRIFTING STYLISTICALLY.  

WHEN A VISUAL DIRECTION IS STILL BEING DEFINED, EARLY CONCEPTS ARE EXPLORED THROUGH RAPID GENERATIVE SKETCHING BEFORE BEING REFINED AND ALIGNED WITH BRAND INTENT. ONCE A DIRECTION IS ESTABLISHED, DATASETS ARE ASSEMBLED AND TRAINED TO ENCODE VISUAL RULES INTO A CONTROLLABLE GENERATIVE SYSTEM.

FINAL IMAGERY IS PRODUCED THROUGH CUSTOM COMFYUI WORKFLOWS THAT MANAGE CONDITIONING, VARIATION, AND OUTPUT STRUCTURE. THIS ALLOWS IMAGES TO BE GENERATED AS COHESIVE SETS RATHER THAN ISOLATED FRAMES, MAINTAINING A UNIFIED VISUAL LANGUAGE ACROSS AN ENTIRE SERIES OR CAMPAIGN.

THE FULL-STACK PIPELINE INTEGRATES MULTIMODAL SYNTHESIS BY PROCESSING THE VISUAL STACK THROUGH KIVE.AI TO ENSURE MOTION CONSISTENCY. THIS IS PAIRED WITH ORIGINAL AUDIO SCORES ENGINEERED VIA UDIO, MATHEMATICALLY SYNCED TO THE METRIC RHYTHM OF THE VIDEO TO ENSURE A UNIFIED AUDIO/VISUAL IDENTIT

I’M MATT GIORDANO — A CREATIVE TECHNOLOGIST BUILDING UNIFIED GENERATIVE SYSTEMS ACROSS FASHION, MOTION, AND AI/ML. MY PRACTICE PRIORITIZES CONTROL, CONSISTENCY, AND AUTHORSHIP THROUGH CUSTOM ARCHITECTED WORKFLOWS.

I TRAIN CUSTOM LORA MODELS USING HEADLESS PYTHON AND PYTORCH, BYPASSING STANDARD INTERFACES TO DIRECTLY MANIPULATE THE TRAINING LOOP. BY SOURCING DATA FROM BRAND ARCHIVES, I DEPLOY MODELS ON PRIVATE GPU INFRASTRUCTURE AND RUNPOD CLUSTERS TO ENSURE SILHOUETTE AND MATERIAL STABILITY WITHOUT STYLISTIC DRIFT.

PRODUCTION OUTPUT IS DRIVEN BY ADVANCED COMFYUI PIPELINES AND PURE PYTORCH INFERENCE. I UTILIZE CUSTOM AUTOMATION SCRIPTING TO ACCELERATE PRODUCTION VELOCITY AND REDUCE WORK-TIME LATENCY, LEVERAGING PROGRAMMATIC CONDITIONING TO ENSURE EVERY FRAME ADHERES TO A UNIFIED VISUAL LANGUAGE.

MOTION AND AUDIO ARE HANDLED VIA KIVE.AI AND UDIO, UTILIZING DATA-DRIVEN SYNCING TO ALIGN THE VISUAL PULSE AND SOUND DESIGN INTO TIGHTLY INTEGRATED MULTISENSORY OUTPUTS. THE RESULT IS A SCALABLE PIPELINE — FROM CONCEPT TO PRODUCTION-GRADE DELIVERY — WITH FULL TECHNICAL CONTROL.

Two stylish young women walking toward the camera at a crowded soccer stadium during a match, with the field and spectators in the background.

Creative Director · Visual Storytelling · Fashion · Image-Making · Campaigns · Creative Technologist · Generative Systems · Editorial · Culture

Creative Director · Visual Storytelling · Fashion · Image-Making · Campaigns · Creative Technologist · Generative Systems · Editorial · Culture

DSKR CREATIVE
DSKR CREATIVE
DSKR CREATIVE
DSKR CREATIVE
DSKR CREATIVE

GENERATIVE AI WORK=))))

GENERATIVE AI WORK=))))

GENERATIVE AI WORK=))))

Neon sign with the word 'DASK' repeated three times in cursive, glowing white and blue against a dark background.
A person wearing sunglasses, a pink headband, a pink graphic t-shirt, a silver jacket, camouflage pants, and light-colored sneakers, kneeling on a grassy hill with an overcast sky in the background.
Young woman standing in a stadium with a soccer field in the background. She is wearing a pink headband, a patterned jacket, and metallic pants.
Young man wearing a black shiny jacket with white text and designs, standing outdoors with a gray wall and buildings in the background.
Young man standing outdoors wearing a black leather puffer jacket with hood, patterned hoodie underneath, and beige pants.
Young person wearing a black shiny Puffer jacket with a hood over a beige hoodie, against a blue background.
A man sitting on a concrete bench, wearing a blue sweatshirt with a black graphic, blue and white camo pants, beige sneakers, and gray socks, holding cash and a white bag with a black graphic
CASE STUDY: SPATIAL & BRAND PERSISTENCE
THIS SERIES DEMONSTRATES A LOCKED GENERATIVE ENVIRONMENT WHERE RETAIL ARCHITECTURE—INCLUDING SPECIFIC SIGNAGE LIKE HOT TOPIC AND SEARS—REMAINS SPATIALLY CONSISTENT ACROSS VARIABLE CAMERA ANGLES.
CUSTOM-TRAINED LORAS STABILIZE DENIM PATCHWORK TEXTURES AND SILHOUETTES, ENSURING THE CHARACTER’S WARDROBE "ANCHORS" TO THE 3D SPACE RATHER THAN DRIFTING STYLISTICALLY BETWEEN FRAMES.
SYSTEM ANALYSIS: MULTIMODAL A/V SYNTHESIS
THIS WORKFLOW UTILIZES LOCALLY TRAINED LORAS DEPLOYED ON PRIVATE GPU CLUSTERS TO RENDER HIGH-FIDELITY ASSETS WITHIN COMFYUI. THE VISUAL STACK IS PROCESSED THROUGH KIVE.AI TO ENSURE MOTION CONSISTENCY.
THE AUDIO IDENTITY IS AN ORIGINAL SCORE GENERATED VIA UDIO, ENGINEERED TO MATCH THE METRIC RHYTHM OF THE VIDEO. THIS DEMONSTRATES A FULL-STACK PIPELINE WHERE SOUND AND VISUALS ARE SYNCHRONIZED.
A young person with long dark hair wearing a black hooded puffer jacket with white lettering, leopard print hoodie underneath, and beige and black patterned shorts, stands in front of a worn, multi-story apartment building with balconies.
Close-up of a person's lower leg wearing brown pants, gray sneaker, and gray sock against a blue background with white text listing city names.

Real-time visual systems for generative AI

Real-time visual systems for generative AI

cables.gl / real-time interaction

An AI-generated image reconstructed as a real-time system rather than a static asset.

We treated the original imagery as source material and rebuilt it inside cables.gl, introducing a GPU-driven deformation layer that turns the image into a responsive surface.

Cursor movement applies localized displacement across the image plane, producing a controlled wave response with natural falloff. The result feels physical, continuous, and deliberate — more material than effect.

The focus is restraint: minimal input, maximum clarity. AI generates the image, but the value emerges through behavior, system design, and interaction.

Matthew Giordano is an editor and creative technologist working in large-scale media production. He spent several years as a staff editor in broadcast post-production at Viacom, supporting network-level content across MTV, Nickelodeon, Comedy Central, and BET.

In high-volume production environments, he focused on pacing, consistency, and delivery at scale. He was also an early participant in Instagram’s Spark AR, creating custom filters that reached millions of users.

Today, through DSKR Creative, he builds generative AI systems for production, designing repeatable image pipelines and translating editorial intent into controllable outputs. The work is grounded in consistency, iteration, and delivery at scale.

CONTACT:
mg@dskrcreative.com

VIACOM CBS

VIACOM CBS

Matthew Giordano is an editor and creative technologist working in large-scale media production. He spent several years as a staff editor in broadcast post-production at Viacom, supporting network-level content across MTV, Nickelodeon, Comedy Central, and BET.

In high-volume production environments, his work centered on pacing, consistency, and delivery at scale. He was also an early participant in Instagram’s Spark AR platform, creating custom filters that reached millions of users.

Today, through DSKR Creative, he designs and builds generative AI systems for production — developing repeatable image pipelines and translating editorial intent into controllable, reliable outputs. His work is grounded in consistency, iteration, and real-world delivery.

CONTACT: MG@DSKRCREATIVE.COM

CONTACT:
mg@dskrcreative.com

MG@DSKRCREATIVE.Com

MG@DSKRCREATIVE.Com