Recently Microsoft released the open 3D generation software Trellis. I decided to try a few tests as reviews had been quite favourable with results creating 3D models with “AI”.
The website describes the technology as unified Structured LATent (SLAT) representation and there is a demo page to try it out. https://trellis3d.github.io/
I chose an image-to-3D example, meaning I upload an image and get back a textured 3D model. Having been a long-time fan of photogrammetry, or creating geometry from two dimentional images, and also a 3D genralist that regularly creates 3D models using Blender, I was eager to see how it would perform.
Using the prompt example seen on VFX & Chill live stream where two artists used Z-Brush to sculpt food in a 20 minute timed challenge, I set out to create a seasonal gingerbread model.
First, Adobe’s Firefly image generation created this image using the prompt: “elaborate gingerbread house with icing“. Impressive start.
Next, dragging the image to the demo site and clicking “Generate” with default settings impressively separated the image from the background and in about 30 seconds showed what it thought the full model would look like. The entire unseen back of the house is amazingly “imagined”.
What’s produced is a GLB file that contains mesh and texture that can be dragged into your 3D program and edited or used as any other asset. In about 4 minutes I have any model loaded ready to use.
As for the model quality, it created a 52K face mesh count which could be considered medium quality. There are lots of artifacts and non-manifold holes but for any background shot it might be just enough to pass the test. Some retopology is almost certainly required and honestly given the choice to “fix” this model or start from scratch I might be tempted to just build it again.
Great from far, far from great.
I don’t give it long, maybe two years, before anyone can drop any fully 3D asset they can think of into any scene realtime.
As with any AI/ML generative model please understand the environmental and ethical implications.
In the autumn of 1892, two miners, Earl and Edward, found themselves high in the Pacific Northwest mountains, searching for a legend as old as the hills: the lost crate of gold. The tale of “Shining Creek Gold” was whispered in every saloon from Seattle to San Francisco, a cache of gold bars rumored to be hidden beneath an ancient cedar tree.
Earl, rugged and quick-tempered, squinted at the worn, hand-drawn map in his hand. Edward, the quieter of the two, adjusted his broad-brimmed hat and gazed over the valley, where fiery autumn leaves painted the landscape.
“We’re close, Ed,” Earl grunted, his voice hoarse from the chilly mountain air. “This map’s leading us right up that ridge.”
As they rode up the narrow trail, their horses’ hooves echoed against the mountain rocks, amplifying the silence around them. The forest grew denser, and the shadows longer, as if the land itself was hiding secrets.
Just before dusk, they spotted it—a massive cedar tree towering over the ridge, its trunk thick as any they’d ever seen. At its roots lay an ancient crate, cracked and splintered, the wood moss-covered and decaying. Inside, stacked neatly, were gleaming bars of gold, catching the last rays of sunlight.
Earl’s eyes lit up, but he kept his voice low. “Look at that, Ed. All the riches we dreamed of.” He grinned and took a step toward the crate.
But Edward didn’t move. His eyes were fixed on something beside the crate—a weathered shovel and, half-buried under fallen leaves, a bleached skull. He crouched down, examining it. The skull’s fractured surface suggested a violent end.
“Looks like someone beat us to it… and paid the price,” Edward murmured.
Earl chuckled, but his laughter had an edge. “Maybe they didn’t know when to quit.”
Edward glanced up, noting Earl’s hand resting just a little too close to his pistol. The air grew tense, thick with unspoken words and simmering greed.
As night fell, the two men set up camp by the tree. They barely spoke, each stealing glances at the crate, the gold gleaming in the firelight. The shadow of the cedar stretched over them, as if watching, waiting.
The next morning, Edward awoke to find Earl standing over him, a grim smile on his face.
“Sorry, Ed,” Earl said, drawing his pistol. “Splitting this gold ain’t in my plans.”
But just as he aimed, a gust of wind rustled the leaves, and a low whisper seemed to carry through the forest. Earl hesitated, looking around as if he’d heard his name. Edward took the moment to roll to the side, grabbing the shovel by the crate.
A shot rang out, but Edward managed to swing the shovel, striking Earl’s arm and sending the gun flying. They grappled by the cedar, the promise of gold twisting their friendship into a deadly game. Suddenly, Earl slipped on loose dirt, tumbling backward, his head colliding with a jagged rock.
Edward stood over him, breathing heavily. Earl lay still, his eyes staring blankly into the sky. Edward let out a shaky breath, realizing what he’d done.
But then he heard it—a faint, rattling breath, coming from Earl’s lips. Earl’s eyes flickered, his hand reaching out, fingers curling around an invisible shovel, as if in some strange delirium.
In his final moments, Earl’s gaze drifted to the skull near the crate, and he muttered, almost as if speaking to it. “Guess I should’ve known… they all come back for their gold.”
Edward shivered, feeling the weight of his actions. He knew he couldn’t take the gold now. It was cursed, a trap that had already claimed more than one life.
As dawn broke, Edward left the mountain, abandoning the crate under the old cedar, untouched. To this day, travelers say they can still hear faint voices near Shining Creek, whispering warnings to those who dare to seek the legendary gold.
And the crate remains, a silent reminder of greed’s deadly grip, hidden beneath the ancient cedar, waiting for the next soul brave—or foolish—enough to try.
Happy Halloween!
Do you know what’s really scary?
This short story was written by a LLM (ChatGPT) Prompt: “Write a clever mystery short story about two miners in the 1890’s. Earl and Edward search for a legendary crate of gold in the mountains of the Pacific Northwest Mountains. A “Who dunnit” style story with a surprise ending.” with images provided by Firefly.
Illustrated by a diffusion model (Firefly) Prompt mostly: “[BLANK] in pacific northwest mountains in fall watercolor scene”
A mixed tape was once a personal show of expression. An arrangement of songs purposefully set to invoke an emotion. Happy, sad, psych-up workouts, breakups, love making… all depends.
Spending time to record a deliberate mix of songs in order for someone else was devotion.
When you’ve heard a specific mixed tape enough you anticipate the next song. Almost like the PBS TV ads promoting Country Music anthologies. One song blends into the next after hearing it so many times you string the next song into the last by memory. (eg. Sixteen tons to Blueberry Hill)
One mixed tape I made, a Maxell Metal 90, for a 90’s cross-country roadtrip journey was burned into my mind. A mix of 60’s to 70’s Neil Young, Led Zeppelin, Canned Heat, etc. and other “hippy” stuff.
I never noticed it went missing but one day, 10+ years later, I was near the Blackcomb Daylodge Rental shop when I heard the transition of one song to the other. I predicted the order of the next three songs, seemingly psychic, only to discover I made the mix a decade before.
Never did I think I was creating something. Not my own. Curating or ordering something was not creating but yet personal preference. Still an art.
In this Apple Music/Google Play/Spotify generation I suspect there will be a hint of nostalgia in playing certain songs together.
A Side / B Side – I wonder if “flipping” media at any specifc point should result in different results.
People not algorithms will ultimately create the most personal selection. AI/ML will still take your job. Rock on!
Generative fill in Adobe Photoshop (beta) is a pretty amazing tool to extend images. Background artists, web designers, print layout artists, or anybody who deals with incorrect aspect ratios or odd image dimentions will be thrilled by the output.
The image below represents a photo I took near Iceberg Lake in Whistler BC. Portrait aspect on 1920×1080 canvas.
To extend the image CTRL+SELECT image. Select Inverse – COMMAND+SHIFT+I SELECT > MODIFY > EXPAND – 3px EDIT > Generative Fill…. (No Prompts)
En Voila! a landscape that does not exists. Notice how it recognizes the light source, maintains correct shadows, tones and palette.
Two thirds of the above image was created by AI/ML yet the middle remains completely untouched and my creation. Does this mean the image was “generated by AI”? “Enhanced by AI”? “Extended by AI”? Or just fake?
Prompts allow for text based editing and generation with mixed results. Note that prompts will only add to the selection. In this case outside of the main image here. Ultimately however it finally gets classified, anyone looking to extend beyond an image will be happily surprised with the results of Generative Fill.
“Degenerate Phil” on the other hand… t
Text prompt “Degenerate Phil” creates some nightmare fuel. Nighty-night.
Generative AI is neither good nor evil. It’s up to you, the human in the loop, to decide.
