Communications career since 2012. Five years deep in generative AI. Three hundred-plus models trained. A wall of recognition from the first crop of AI film awards. And a cat with more trophies than I do. I make cutting-edge tech do things it probably wasn't designed for - but should have been.
CALE FROMBACH
LAKE COUNTRY, BC
EST. 2024 → 2026
Plot twist: I went to business school to learn about PowerPoint and spreadsheets, and ended up teaching AI models to make rap videos about cats. Life's funny like that. Communications career since 2012, which mostly meant explaining complex tech without inducing comas — a skill that turns out to be the entire job once AI showed up.
Speaking of which: somewhere around 2023, generative AI broke every assumption I had about what a small team could pull off. I went deep. Then deeper. I specialize in making these tools do things they probably weren't designed for (but should have been). My work ranges from serious business consulting to content that makes people question their life choices in the best possible way. I'm basically a translator between "cutting-edge AI research" and "what the hell does this button do?"
Three hundred-plus published models later, here we are. I've trained FLUX, Stable Diffusion, Qwen, and video models like they're digital pets. Built ComfyUI pipelines that scared me a little when they ran. Won some awards — my cat won more. And got really specific about what this technology is genuinely good at, versus where it's still mostly marketing.
Based in Lake Country, BC, where the WiFi is strong and the coffee is stronger — alongside my partner in crime Allie (the cat) and Lia (the dog). I work with clients globally who appreciate both technical expertise and the occasional dark sense of humour. When I'm not doing this, I've bought-fixed-flipped two foreclosed homes, manage a serious investment portfolio, and recently finished a horticulture certificate. How's that for variety.
Now it's three things: I direct my own work (the videos up top), I teach creative teams the real stack (the training), and I drop into studios that need someone who's actually shipped this in the room (the consulting). That's the whole business. No funnel, no growth-hacking, no BS.
The fastest way to know what to build is to build something embarrassing and look at it. Every project I've ever shipped started ugly.
If someone is selling you a prompt template, they have not shipped anything real. The work is in the iteration, the constraints, and the taste - not the spell.
Every closed AI platform is one pivot away from breaking your workflow. The studios that stay flexible are the ones that own the connective tissue.
The model doesn't know what's good. Whoever owns the taste at the end - the eye that says yes or no on the final frame - is the most important person on the project. <strong>That role doesn't get automated, and pretending otherwise is the trap most studios are walking into.</strong>
Every studio I trust documents how they got to the final shot. The ones that don't usually can't reproduce it - and definitely can't teach it. <strong>Pipelines are public; secret sauce is mostly insecurity.</strong> Breakdowns, ComfyUI graphs, BTS - I share them.
The whole industry is racing to the bottom on token costs and template prompts. The studios that win are spending on the things that don't show up on a feature comparison — *taste, judgment, custom infrastructure.* You get what you pay for, and right now most companies are paying for the work to look like everyone else's.
Went in for PowerPoint and spreadsheets. Left with a communications career and absolutely no idea what was coming.
Years of explaining complex things to people without putting them to sleep. The skill that quietly became the whole job.
A free model, a bad GPU, and a lot of horrifying hands. But also the first time the screen showed me something I'd thought up ten seconds earlier. *Couldn't stop thinking about it.*
A weekend, some broken documentation, and one moment of "*oh.*" The thing that changed the trajectory. Haven't really stopped since.
Left a job that wasn't a good fit and went all in. <strong>Found the rare role that wants both the technical and the creative.</strong> Worked with Civitai, ThinkDiffusion, and Native Foreign - three different angles on the same industry, all of them shipping real work.
Pivot from a portfolio site to the three-thing studio it actually is: direction, training, consulting. This site.
Inaugural training program in active setup — curriculum locked, cohort being scoped. Spots opening for Q3 2026. Book a call to see if your team fits.
Available for one or two new fractional AI lead engagements through 2026. Audits and short projects open more flexibly - those don't take a slot.
Booking commercial video work for Q4 - AI-assisted shoots, music videos, brand pieces. Pipeline conversations open now.
