Get Comfy with LAIA
How Much Can You Build with AI?
Join AI for Architects Starting Sep 22nd
& Find Out!
Hey {{first_name | Friend}} 👋
Adam here, thinking about the insane era we’re blessed to live in — one where technology has advanced beyond what we previously believed possible; one where we’re able to use technology to perform feats we couldn’t have conceived of before.
Yes, amigos, these are thrilling times we’re moving through, and I feel grateful to be on this wild ride with Lighthouse AI Academy: a beacon of direction through the often bewildering and rapidly changing world of AI.
This week, our ComfyUI cohort dove deep into advanced prompting and workflow design, and the timing couldn’t be better.
Simultaneously, NVIDIA unveiled stunning breakthroughs at Gamescom, where artists and modders used AI to reimagine classic games with RTX Remix.
Different worlds, same story: AI isn’t replacing artists, it’s enhancing them.
When humans lead and AI supports, creativity accelerates.
That’s the essence of what we’re teaching: not button-clicking, but a philosophy of workflow design.
As our ComfyUI course director, Maged Elbanna, always reminds us:
Never stare at the mountain.
Break it into islands and make each island work.
👁️ Inside the Cohort: Advanced ComfyUI Q&A
Here’s a snapshot of what we covered, with visuals placed throughout the section so you can follow along with ease.
Week 2 Q&A — advanced workflow design mindset
1. Prompt Structure Matters More Than Ever
A student asked why prompt length/structure changes results so much across SD 1.5, SDXL, and Flux.
Maged explained: there’s no one “rule.” Instead, it’s about systematic design.
Each model responds differently; what matters is thinking programmatically.
How human language maps to model encoding
“You’ll never get a straight math answer from these systems.
You hold a vague outcome in mind and fine-tune your way toward it.
— Maged Elbanna
2. Seeds = Consistency
Setting random seeds leads to endless variation.
Fixing the seed = repeatable results across machines.
Agencies already share PNGs that encode workflows — so peers can recreate identical results.
Sampler vs. scheduler — where denoise lives
3. Cloud vs Local
Local machines choke on video models.
Best practice: start with cloud (RunPod, RunComfy) to test GPUs before investing thousands in 5090s.
Quantized models can reduce generation time but cost some detail.
End-to-end graph:
checkpoint → sampler → conditioning → VAE → preview.
“Don’t spend on a 5090 until you know what you need.
Rent one for an hour and see what you actually get.”
— Maged Elbanna
4. Workflow Methodology
Maged’s personal approach:
Always start from scratch, build the sampler, choose the model.
Generate until a result points in the right direction.
Fix the seed.
Layer islands: inpainting → conditioning → LoRAs → adapters.
Many directions, all with good results
I always build from the ground up. Keep it simple. Break problems into mini-workflows. Connect them and suddenly, you’ve built something complex.
— Maged Elbanna
🧠 Homework: Seeds, Denoise, and Iteration
Your exercises this week:
Rebuild a workflow with randomized vs fixed seeds → compare consistency.
Test Flux denoise values at 70, 90, 100 → notice where control tips.
Share one before/after on LinkedIn → tag Lighthouse AI Academy.
AND REMEMBER:
Iteration = mastery. 🤺
🔬 AI as an Enhancer, Not a Replacer
This week crystallized a truth we’ve carried since day one:
Artists aren’t being automated out of the picture; they’re being augmented.
The breakthroughs come when:
Workflows are modular and repeatable.
Prompting becomes a craft, not a plug-and-play.
AI doesn’t reduce the creator’s role; it reshapes it.
The time once spent wrestling with repetitive setup is now invested in higher-order skills: curating prompts, architecting workflows, and refining concepts into systems.
That’s why we’re asking our students to begin journaling their journey at Lighthouse AI Academy, capturing the prompts, workflows, and reflections that mark their growth.
And if you’re reading this as a subscriber, you’re invited too.
Share your experiments, your wins, or even your struggles. We’d love to feature and collaborate with all AI creators, because every perspective expands what’s possible.
We should run an open-door policy. If you have something to share, whether work, developments, or experiments, we’ll highlight it across our channels: LinkedIn, articles, and newsletters.
— Nejc Susec
That’s why we run cohorts and publish these newsletters: to show how AI expands the creative canvas without replacing the artist’s vision.
🔦 Industry Spotlight: NVIDIA RTX Remix + AI Blueprints
At Gamescom, NVIDIA proved just how powerful this idea is:
Need for Speed: Underground RTX Remix → 500+ textures generated with ComfyUI, then polished by hand.
RTX Remix reimagines “Need for Speed: Underground’s” Chinatown: RTX OFF (left) shows low-resolution textures and flat lighting, while RTX ON (right) adds neon reflections, detailed materials and fully path-traced streets.
Unreal & Portal 2 Remixes → small mod teams pulling off studio-level overhauls in months, not years.
RTX Remix transforms the alien environments in “Unreal”: RTX OFF (left) shows the original flat look, while RTX ON (right) adds detailed PBR materials, emissive lighting and realistic reflections.
Painkiller RTX Remix → 35 levels remastered with PBRFusion + RTX GPUs.
RTX Remix transforms the gothic cathedral in “Painkiller”: RTX OFF (left) shows the original look, while RTX ON (right) fills the hall with stained glass reflections, volumetric beams and realistic shadows.
Meanwhile, the new NVIDIA AI Blueprints let 3D artists generate up to 20 draft objects from a single prompt, then refine them instantly in Blender.
As this game creator put it:
“Generative AI was mainly used for texture generation. It’s impossible to create such a large number of textures in such a short period of time alone without AI.”
Once again, AI didn’t replace artistry: It freed it.
📖 New Blog Post — Move Your Bones: Reimagining Music Video Production With AI
We’ve just published a new blog post that connects directly to what we’ve been exploring together:
The piece unpacks how AI is reshaping music video production, turning traditional workflows into dynamic, experimental pipelines.
Watch the full music video here.
The parallels with ComfyUI and NVIDIA are most clear:
Just as RTX Remix frees game modders to focus on artistry, ComfyUI workflows free video creators to push into new territories.
What we’re building in our cohorts—prompting, iteration, modular workflows—maps directly onto the way artists are experimenting with AI in music, film, and beyond.
The boundary isn’t AI vs. artist. The boundary is how far we’re willing to push.
In short: what we can do with Comfy is limitless.
➡️ Coming Up Next
Q&A Masterclass Friday for ComfyUI cohort→ covering core topics with hands-on project work.
Creative Leaders Program → launches Sept 9.
Stay tuned—we’ll be drawing even tighter links between cohort insights and real-world case studies.
Upcoming Cohorts: Enrol Now
👷 AI for Architects — Sept 22–Dec 12, 2025
Turn sketches into cinematic master plans.
Concept-to-visualization pipeline.
Signature style via LoRA training.
A peer network shaping the next frontier.
Who It’s For:
Architects, viz specialists, urban planners, studio owners, computational designers, and creative technologists.
Outcomes:
Portfolio-ready project
Faster iterations with better control
A network of peers pushing the field forward
👉 Apply for AI for Architects before seats run out!
🤹 AI for Creative Leaders — Launched on Sept 9
Lead teams to design repeatable, commercial AI workflows.
👉 AI for Creative Leaders is waiting for you.
🌄 Closing Sentiments
This week reminded us:
AI isn’t the artist. YOU are.
And with the right workflows, AI gives you more time for vision, less for busywork.
That’s the real breakthrough, and it’s only just beginning.
If you’re serious about mastering AI, our cohorts are designed to help you do just that; we’d love to have you join us. 🔥
That’s it for now: Thanks for reading and building this new era with us.
There is a lot happening in the world of AI.
ALL. THE. TIME.
That means there’s constant room for creation, too. At our academy, we encourage and can’t wait for more creatives to join us as we ride the diffused wave into the next era.
It’s as easy as signing up with us today. When you’re ready:
Step by step, node by node — and away we go!
Keep creating and always remember to have fun.
— Adam & the Lighthouse AI Academy Team ☀️
Small Team, Big Dreams