AI & Marketing

Training Creativity — How AI Learns to See, Imagine, and Create

By Ilya Mezerovsky

Introduction: Who Am I?

• Background in as developer and product creative & interactive products.
• Joined monday 3.5 years ago.
• AI Creative and Productive Advisor at monday.com.
• Worked on a variety of projects and product in monday.
• Been using, exploring and developing AI based tools before they were cool.

Introduction: What I do at monday.com?

• Help teams at monday to identify and define and develop impactful AI workflows and use cases.
• Research and prototype new AI tools and models.
• Mentor teams on AI best practices for their existing tools.

My AI Involvement

• Depth-based Prototypes – Combined image generation with depth estimation
  (Video demo)
• AI Worlds – AI video generation, segmentation, and eye-tracking
  (LinkedIn Post)
• Using AI for any conceivable use-case
  • Even attempted to create this presentation with AI tools (badly)

Gen AI basic terms and concepts

AI model in simple terms

Generative AI model is a system trained on huge datasets to learn patterns and then create new content - text, images, code, audio, etc. that resembles what it has seen. 

Closed Source vs. Open Source AI

• Accessibility
• Control & Customization
• Licensing Concerns

Gen AI Lifecycle

Gen AI models usually have 2 distinct stages in their lifecycle

1. Training & Fine‑Tuning - using a datasets to train model weights.
2. Inference - running the trained model to generate output.

Best use-cases for GenAI

• Automating repetitive tasks
• Streamlining workflows
• Improving decision-making

Closed Source vs. Open Source AI

• Closed Source AI

• More stable, polished, and tested solutions
• Stronger security & controlled access
• Official customer support & documentation
• Doesn't require hardware or cloud investement

Closed Source vs. Open Source AI

• Open Source AI

• Transparent & customizable
• Faster innovation with community contributions
• Cost-effective with no licensing fees
• Greater flexibility & adaptability
• Can be run locally

Popular Gen-AI domains and tools

Popular Gen-AI domains and tools

• Natural‑Language Generation – ChatGPT, Claude, Gemini

• Image Synthesis & Editing – 4o, Midjourney, Stable Diffusion, Flux

• Video Creation & Animation – Runway, Pika, Sora, Veo 2, WAN 2.1

• Audio & Music Production – Suno, ElevenLabs, MusicGen

• 3D Asset & Scene Generation – Hunyuan 2.5, LumaAI, Meshy, TRELLIS

• Code & Dev Assistance – Cursor, Base44

monday mansion usecase study.

What is monday mansion?

Our venues.

Defining our need:

Creating 3D venue mocks manually is expensive and time consuming, we want to speed up the process.

First Attempt: Prompt Engineering

"Show me a picture of lots of blooming California poppies, make them bright, vibrant orange, and draw them in an illustrated style with colored pencils"

"A bad prompt example" - Midjourney documentation

Live Experiment

Can you create a prompt to describe "Starry Night" without

mentioning the artist, style or painting name?

Live Experiment

Can you create a prompt to describe "Starry Night" without

mentioning the artist, style or painting name?

Using ChatGPT to help

"A vibrant, expressive painting that depicts a swirling night sky filled with glowing orbs resembling stars. The composition uses bold, visible brushstrokes that create a sense of movement and energy. The colors are rich and saturated, dominated by deep blues, with bright yellows and whites forming the celestial elements. There is a strong contrast between the cool, dark background and the luminous, warm highlights. In the foreground, elongated, dark organic shapes rise toward the sky, adding depth and a dynamic vertical element. The brushwork is loose and textured, emphasizing emotion and atmosphere rather than strict realism. The scene conveys a dreamlike, almost surreal quality, as if capturing the essence of a vivid, imaginative night rather than a literal representation."

First Attempt: Prompt Engineering mansion venues

Trying to Prompt Engineer mansion venues

• A smaller, specialized “add-on” for large models

1. Define the Topic (e.g., Mansion Rooms)
2. Gather & Prepare Data (labeling, etc.)
3. Train the LoRA
4. Test & Iterate

Building a Custom LoRA

1. Define the Topic (e.g., Mansion Rooms)

Building a Custom LoRA

2. Gather & Prepare Data (labeling, etc.)

Building a Custom LoRA

2. Gather & Prepare Data (labeling, etc.)

1. 3-300 images – for style loras try varying objects and scenarios.
2. Keep the images in the same aspect ratio.
3. Iterate Often – Refine data & prompts continuously

Building a Custom LoRA

Building a Custom LoRA

4. Test results

Building a Custom LoRA

4. Test results

Building a Custom LoRA

4. Test results

Building a Custom LoRA

4. Test results

Building a Custom LoRA

4. Test results

Building a Custom LoRA

4. Test results

Building a Custom LoRA

4. Test results

Business Impact

• Faster Content Creation
• Reduced Costs & Time
• Competitive Edge via unique brand visuals

Thank You!

Q&A / Discussion

Thank You!

Q&A / Discussion

Resources and

Use-cases

Text Gen AI

Use Cases:

• Content Generation: Articles, blog posts, marketing copy, product descriptions.

• Conversational AI: Chatbots, virtual assistants, automated customer support.

• Code Generation: Autocompletion, code refactoring, documentation generation.

• Summarization: News aggregation, meeting minutes, legal document summarization.

• Translation & Localization: Language translation, dialect adaptation, cultural customization.

• Creative Writing: Storytelling, poetry, scriptwriting.

• Text-based Game/NPC Dialogue: Dynamic character interactions, procedural quest generation.

• Data Analysis & Insights: Report automation, financial analysis, trend prediction.

Text Gen AI

• ChatGPT (OpenAI) – A conversational assistant
• LLaMA (Meta) – An open-source language model
• Claude (Anthropic) – A writing & project management assistant
• Custom Finetunes – Tailored open-source models trained on specific data

Image Gen AI

Use Cases:

• Image Generation: Concept art, marketing visuals, social media graphics.
• Inpainting & Editing: Object removal, image enhancement, style transfer.
• Super-resolution: Improving image quality (e.g., upscaling old photos).
• AI-driven Design: Logos, UI/UX elements, 3D textures.
• Product Visualization: E-commerce previews, fashion and interior design mockups.
• Medical Imaging Analysis: AI-assisted X-ray, MRI, and CT scan analysis.
• Security & Surveillance: Face recognition, object detection.

Image Gen AI

• Midjourney – Text-to-image creation
• Stable Diffusion – Open-source text-to-image
• Flux – Photo-realistic image generation

Video Gen AI

Use Cases:

• Video Generation: AI-generated films, animated explainer videos, deepfake synthesis.
• Video Editing & Enhancement: Auto-colorization, frame interpolation, stabilization.
• Virtual Production: AI-assisted CGI, background replacement, motion capture.
• Personalized Content: AI-driven dynamic ads, customized video messages.
• Automated Video Summarization: Highlight reels, scene detection for content creators.
• Security & Surveillance: AI-powered threat detection, anomaly tracking.
• Synthetic Training Data: AI-generated simulations for machine learning training.

Video Gen AI

• Runway – AI-powered video creation & editing
• AnimateDiff – Text-to-animation approach
• Hunyuan - Opensource video generation

3D Gen AI

Use Cases:

• 3D Asset Generation: Game development, architecture, digital twin simulations.
• Procedural Environment Creation: AI-assisted world-building in games & VR.
• 3D Character Creation: AI-generated humanoids, creatures, avatars.
• Motion Capture & Animation: AI-driven pose estimation, facial animation.
• Manufacturing & Prototyping: AI-generated CAD models, 3D printing optimization.
• Medical Applications: AI-assisted prosthetic design, anatomical modeling.
• Augmented Reality (AR) & Virtual Reality (VR): AI-powered object recognition and placement.

3D Gen AI

• Trellis – AI-driven 3D modeling
• Meshy – Automated 3D mesh generation & optimization
• HunYuan - Open sourced 3D models 

Audio Gen AI

Use Cases:

• Speech Synthesis & Voice Cloning: AI-generated voiceovers, personalized digital assistants.
• Music Generation & Enhancement: AI-composed music, auto-mixing, and mastering.
• Sound Design & Foley: AI-generated sound effects for movies and games.
• Noise Reduction & Enhancement: Podcast cleanup, audio restoration.
• Speech Recognition & Transcription: AI-driven subtitles, accessibility tools.
• Emotion & Sentiment Analysis: AI-powered call center analytics, audience sentiment tracking.

Audio Gen AI

• Suno – AI-generated music
• Whisper (OpenAI) – Speech-to-text
• ElevenLabs – AI voice synthesis & cloning