Behind FrameArto: The Story, The Craft, The Promise
FrameArto is a small studio building custom AI portraits the slow way, with prompt engineering tuned to each style, three-pass generation, and a real person looking at every print before it ships. Here is who we are, what we believe, and why we built it.
Matt Morgan
Founder, FrameArto
Hi, I am Matt. Here is why I built this.
In late 2023 I ordered a custom portrait of my parents' old labrador from a competitor for what was supposed to be a Christmas present for my mum. It cost me sixty pounds, took eleven days to arrive, and when I opened the box I almost cried, not because it was beautiful, but because the dog in the painting did not look anything like the dog I had grown up with. Wrong eyes, wrong coat, wrong expression. I had no recourse. I had paid. I had received a thing. The thing was not what I wanted. I sent my mum flowers instead and the painting went in a drawer.
A month later I was reading about new diffusion models that had just come out and it struck me that the technology to make a far better portrait of that labrador now existed, openly, for any developer to use. The reason my sixty-pound portrait had been bad was not that AI was incapable. It was that the company I had bought from had built a fully automated pipeline with no human in it anywhere, and shipped whatever the model produced. The craft layer was missing. The care was missing.
FrameArto started a few months later as a side project to prove a point: that you could use AI seriously as a tool and still treat each order like a commission. We do prompt engineering per style, generate three variations, look at every canvas with human eyes before it ships, and run an actual support inbox staffed by actual humans. That is the whole pitch. The models will keep getting better. The craft layer is what we own.
Six principles, one studio
Everything we ship, every decision we make about pricing, support, and craft, comes back to these.
Every portrait should feel like a gift
Even when you are buying it for yourself. From the moment the canvas arrives in your hands to the moment it is on your wall, the whole experience should feel considered. Packaging matters. Presentation matters. The unboxing matters.
AI is a tool, craft is the differentiator
Anyone can run a photo through a generic model. What separates a forgettable AI image from a portrait you hang on the wall is the craft layered on top, the prompt engineering, the iteration, the human eye that catches what the model missed.
Cheap should not mean low quality
A $9.95 digital download should not look $9.95. Pricing low does not have to mean cutting corners on output, packaging, or care. It just means we have built efficient operations behind the scenes so the quality stays high.
If it is wrong, we fix it
Every order includes three free retries because no AI gets it right every time. If the third attempt still misses, we reach out personally. Refunds are not a fight. The goal is a portrait you actually want, not a transaction we closed.
No portrait ships unseen
Before any canvas leaves the print partner, a human looks at it. If the AI dropped a detail, if a hand looks wrong, if a colour is off, we catch it and rerun before you ever see it. This is why we are slower than fully-automated services. It is also why we are better.
Your photos are your photos
We do not train models on your uploads. We do not resell your images. Your photo is used to generate your portrait, then archived for thirty days in case you want to re-order, then deleted. That is the whole policy.
A four-step process, not a black box
This is what actually happens between your upload and the canvas arriving on your doorstep.
Photo intake and review
You upload a photo through the preview tool. Behind the scenes, the image is checked for resolution, framing, and lighting. If the photo is going to cause problems (too dark, too low-resolution, faces obscured), the system flags it before we waste a generation cycle.
Style-specific prompt engineering
Every style on FrameArto has its own dedicated prompt template, refined over hundreds of test generations. A watercolour prompt is not just an oil painting prompt with the word "watercolour" swapped in. Each one has its own lighting cues, brush descriptors, composition notes, and negative prompts to keep the model on style.
Three-pass generation
You receive three variations from a single upload, generated in parallel with intentionally different prompt seeds. This is not just to give you a choice. It is also our quality net. If two of three come back wrong, that tells us something about the source photo or the prompt and we adjust before the retry.
Human-eye QA before delivery
For every canvas order, a real person reviews the file before it goes to the print partner. We are looking for the things AI still gets wrong, fingers, jewellery, fur direction on pets, the symmetry of glasses, the small details that separate a good portrait from a great one. If it does not pass, we rerun. You do not see the misses.
Plain English, no fine print
Buying art on the internet from people you have never met is a leap of faith. We know that. So here is exactly what we owe you, in language you can actually read. No legalese, no escape hatches, no surprises.
If you have a question about any of this before ordering, email support@framearto.com. A person will reply.
Free preview, no card required
You only pay when you decide you like what you see. No deposit. No "purchase to unlock". The three preview variations are free, full stop.
Three free retries on every order
If the first generation does not land, you can regenerate up to three more times at no cost. Most customers find their portrait on the first or second pass.
Seven-day return window on prints
If your canvas arrives and it is not what you wanted, ship it back within seven days for a full refund. No restocking fee. No back and forth.
Damaged in transit replaced free
If your canvas arrives bent, scratched, or otherwise damaged by the courier, send a photo to support and a replacement ships within 48 hours. No questions.
Real human support, not a chatbot
Email support@framearto.com and a person replies. Most responses go out within four working hours, often faster. If you need to talk to someone, you talk to someone.
The honest, slightly technical bit
Under the hood, FrameArto runs on a stack of current-generation diffusion and image-to-image models, including Flux Pro for fine-detail likeness work, SDXL for style-heavy variants like cartoon and pop art, and a custom pipeline that combines them where neither alone gets the job done (renaissance oils, for instance, use a two-stage pass to get the chiaroscuro right).
Each style on the site has its own dedicated prompt template, lighting profile, and negative prompt set. These are not generic. The watercolour template was built by running roughly four hundred test generations against reference plates of the actual medium, then refining the descriptors that produced the closest match. The renaissance oil template was built the same way against Rembrandt and Vermeer references. We do this for every style we ship.
New styles get added only when we can pass our own internal “would I hang this on my wall” test across a representative range of source photos (different skin tones, different lighting conditions, pets, kids, multi-person compositions). If a style does not pass that bar across the board, it does not ship. There are styles we have tested and shelved because the output was not consistent enough. That filtering is why the list of available styles grows slowly. We would rather offer thirty styles that work than a hundred that mostly do.
As featured in
Real people, real portraits
Three of the customers who took the time to tell us what their FrameArto portrait meant to them. In their words, with their names.
Ready to see what we can do?
Free preview, no card required. You only pay when you find a portrait worth keeping.