Hey {{first_name | Friend}} 👋

Adam here with another student project that stopped us in our tracks: finnessed, restrained, and technically sharp in all the right ways.

You’ve seen AI used for spectacle.

You’ve seen it used for speed.

But this week’s project asks a different question:

Can AI be used to create stillness?

That’s exactly what Lighthouse AI Academy student Dennys Herman set out to explore with Temple Moss, a cinematic fragrance film that feels less like a “generated video” and more like a high-end perfume campaign you’d expect from a luxury brand studio.

This is a project about discipline, restraint, and knowing when not to push the tools too far.

Ready? Let’s get stuck in!

Temple Moss is a short cinematic project that blends fragrance, film, and visual poetry.

The goal wasn’t to explain a product; it was to evoke a memory.

Inspired by abandoned temples slowly reclaimed by nature, the project explores stillness, decay, and atmosphere. The guiding idea was simple but demanding: create something that looks practical, tactile, and intentional — not algorithmic.

Rather than leaning into spectacle, Temple Moss aims for high-fidelity abstraction, mirroring the visual language of luxury perfume campaigns while quietly testing the real limits of current AI tools.

Before opening any software, the work began with the hardest part: defining something you can’t see.

The fragrance concept was titled “Nature Remembering.”

Scent profile:

Each note became a visual rule:

This wasn’t about illustration — it was about translation.

One of the biggest challenges in AI video is consistency across time. Shots generated days apart often feel like they come from different cameras, or different worlds.

To solve this, Temple Moss built Custom GPTs specifically for prompt generation.

These models were conditioned on a tightly curated reference set that defined:

The result? Ideas could stay loose, while execution stayed strict. Concepts were filtered through a predefined visual identity before ever hitting an image or video model.

This entire system was built before newer model updates simplified the process, making it a true example of workflow design over tool reliance.

Instead of chasing perfect compositions, the project prioritized texture.

Stone had to feel ancient, porous, weathered, quiet.
Moss had to feel soft, damp, and alive.
Smoke had to feel like incense, not fog.

For the bottle itself, the approach shifted again.

Glass is notoriously difficult in AI. When generated in-environment, refraction artifacts and blending errors destroy realism. The solution was to treat the bottle like a studio product shoot: isolated, controlled, and grounded in physical logic.

Heavy glass. Cold surfaces. Condensation. Neutral voids.

This separation preserved material realism and avoided the “warped plastic” look that often gives AI away.

The visuals demanded restraint and so did the sound.

Instead of a melodic score, the audio was treated as environmental texture:

Using ElevenLabs, sound was layered like foley rather than composed like music. The narration itself required dozens of generations, stitched together from multiple takes to achieve the right cadence and calm.

Stillness, it turns out, is expensive.

Originally, the vision called for seamless morphing transitions, stone dissolving into smoke, smoke becoming glass.

In practice, the models couldn’t handle the complexity without hallucinations, artifacts, or massive fidelity loss. Iteration burned credits fast, and quality dropped even faster.

The solution was pragmatic: abandon the AI morphs.

By returning to traditional cinematic cuts and slow fades in DaVinci Resolve, the project regained control over pacing and realism. Editing became curation, finding the few clean seconds of motion before reality broke.

The takeaway was clear:

If the tool fights you, pivot.

Temple Moss isn’t about pushing AI harder.

It’s about using it more carefully.

The project shows that cinematic quality doesn’t come from complexity — it comes from systems, taste, and restraint. AI becomes most powerful when paired with traditional craft, not when replacing it.

This is exactly the mindset we teach at Lighthouse AI Academy.

If this workflow resonates with you or you have any further questions, please reach out.

We would love to get your feedback on what works and what doesn't regarding this process.

Inspired by Dennys and his Temple Moss project? Try this:

Discipline is a creative choice.

Feel free to share your outcome with us by email, or tag us on LinkedIn and show us what amazing things you’ve created!

Projects like Temple Moss don’t come from prompts alone; they come from systems thinking.

If you want to build production-ready workflows like this, applications are open for our upcoming courses:

🟡 Basic ComfyUI

 → Build clean, readable visual workflows from scratch — no prior node experience required.

🟡 AI for Creative Leaders

→ Design systems (not just outputs), integrate AI with teams, and lead with strategy and ethics.

🟡 AI for Architects: 2-Week Studio

→ A fast-paced, hands-on studio for architects and designers ready to ship.

👉 View all courses or book a call with our team to see which path fits you best.

Dennys’ Temple Moss reminds us that boundaries are only temporary.

AI wasn’t built to “visualize a scent.” And yet with taste, discipline, and a willingness to pivot when the tools push back, this project proves you can translate something invisible into something cinematic.

This is what happens when creatives show up with restraint, curiosity, and the patience to build systems that protect consistency.

That’s it for now: thanks for reading and for building this new era with us.

If you’re ready to go from experiments to repeatable workflows, applications are open for Basic ComfyUI, AI for Architects, and our AI for Architects Studio, and if you’re unsure what fits, book a call, and we’ll guide you. 💛

Step by step, node by node — and away we go.

Keep creating and always remember to have fun.

— Adam & the Lighthouse AI Academy Team ☀️