Yes, quantum computers can generate images today, and they do it on real hardware. The images they produce are tiny, typically 28 by 28 pixel greyscale handwritten digits from standard research benchmarks, made by hybrid systems in which a quantum circuit generates small patches of the picture and a classical neural network judges the result. A single Instagram post holds roughly 4,400 times more values than one of those digits. Campaign-grade AI images and AI influencers come from classical generative models, and quantum image generation remains a research field to watch.
The short version
- Quantum image generation is real. Quantum GANs run on real processors and produce real pictures
- The pictures are benchmark sized. Usually 28 by 28 greyscale digits or similar small images
- The systems are hybrid. A quantum circuit generates image patches, and a classical network does the judging
- Size is the smaller problem. Generalisation, an analytical fidelity ceiling and vanishing gradients are harder
- Brands should use classical AI today. Diffusion models already deliver campaign-grade imagery
Why the question comes up
The word "quantum" now appears next to "generative AI" in vendor presentations and technology news, and marketers are starting to ask whether the images behind their next campaign might be produced on a quantum computer. It is a reasonable question, and it has a precise answer. Quantum image generation exists, it runs on real hardware and it produces real pictures. What settles the planning question for any brand is what those pictures actually look like.
What a quantum GAN is
A normal generative adversarial network, or GAN, has two halves. A generator invents images, a discriminator judges whether each one looks real, and the two train against each other until the generator succeeds often enough to be useful.
A quantum GAN, or QGAN, replaces the generator with a parameterised quantum circuit. The discriminator stays a classical neural network. That detail matters more than it first appears. Current QGANs are hybrid systems, and the judging half is ordinary machine learning.
There is a second compromise. Quantum processors available today are small, far too small to represent a whole image, so the image is divided into patches. Each patch gets its own sub-generator circuit. Random noise is fed into each one, the output of each circuit is mapped onto pixel values, and the patches are stitched back into a single picture.
What has actually been built
The clearest working example comes from the photonic quantum company Quandela, whose QGAN algorithm generates artificial images from existing image datasets, implemented on its photonic processor and trained end to end. The design follows the patch-based approach described above, with noise feeding variational photonic circuits, sub-generator outputs mapped to pixels, recombination into a finished image and a classical neural network as the discriminator. The implementation has been published openly.
The stated commercial purpose is synthetic training data. Generating artificial images for many scenarios can reduce the cost and time of collecting specialised image datasets and increase the variety of data used to train other AI models. That matters in sectors such as automotive and aerospace, where real footage of rare edge cases is expensive or dangerous to collect.
That is a genuine and sensible use case. It is also a very different job from producing a campaign image, and the difference is where brands get confused.
The output, in numbers
Published quantum image generation works at benchmark scale. As described by Quandela and ZKSF.org, whose practical overview of quantum machine learning on simulators explains why this training is so expensive, the standard datasets are the ones the whole field uses.
| What is generated | Size | Values per image |
|---|---|---|
| MNIST handwritten digit, the usual QGAN benchmark | 28 x 28 greyscale | 784 |
| Fashion-MNIST clothing item | 28 x 28 greyscale | 784 |
| CIFAR-10 object | 32 x 32 colour | 3,072 |
| One Instagram post | 1080 x 1080 colour | 3,499,200 |
| One 4K portrait | 3840 x 2160 colour | 24,883,200 |
A single social post carries roughly 4,400 times more values than one MNIST digit. A 4K portrait carries about 31,700 times more. That is the real scale of the distance, and it will take far more than one hardware generation to close.
Why resolution is the smaller problem
If image size were the only obstacle, the field could simply wait for larger processors. Three deeper problems make the gap harder to close.
Generalisation. Published work finds that QGANs struggle to generalise across datasets and tend to converge on the average representation of their training data. For a brand, that is a serious weakness. An AI influencer campaign needs variety, distinctiveness and a consistent character, and a model that collapses toward the average of its training set produces the blandest possible output.
An analytical ceiling. When the generator's output is a pure quantum state, the achievable discriminator quality is bounded by the fidelity between that state and the target data distribution. This is a mathematical limit on the architecture, so better hardware alone does not remove it.
Training cost and vanishing gradients. Quantum machine learning needs thousands of circuit evaluations per training run, and variational circuits of this kind run into barren plateaus, regions where the gradient effectively vanishes and the optimiser stops learning. Anyone who has trained a classical model knows what a dead gradient costs. In these circuits the problem tends to grow as the circuits get wider.
Why vendor pages rarely show these numbers
In fairness to vendors, a use case page exists to describe a capability to a commercial audience, and the underlying research is usually open for anyone who wants the detail. The practical consequence still falls on the buyer. A page describing artificial image generation on a quantum processor rarely mentions that the output is a 28 by 28 digit. Finding that out means reading the research, which most people planning a marketing budget reasonably do not do.
One simple rule protects you. Ask what it produced, at what size and measured how. A claim about quantum capability that cannot answer those three questions describes a roadmap.
How classical AI produces campaign images today
Campaign-grade images and AI influencers come from classical generative models, mainly diffusion models trained on very large image collections and run on ordinary GPUs. They produce photorealistic people, products and scenes at high resolution, and, with the right production system, keep characters consistent across hundreds of images and videos. That is how Proklisi creates SI influencers and campaign imagery for brands. Our guide on creating photorealistic AI images explains the craft, and our explainer on generative AI in advertising covers the technology.
Meet the SI influencers who could front your campaign
What a brand should do about quantum image generation
For campaign planning, the useful answer is to carry on with classical AI and keep an eye on the research.
- Leave quantum creative out of the budget. There is no product to buy for campaign imagery, and the gap is around four orders of magnitude.
- Keep it out of the critical path. Waiting to see what quantum does is no substitute for a content strategy. The timeline is a research timeline.
- Treat it as a real field. Photonic and other processors really are generating images, and the synthetic training data use case is sound. It is simply a different tool from a creative production pipeline.
- Watch two specific milestones. Qubit counts say little on their own. The signs that matter are demonstrations of generalisation beyond the training-set average and generation above benchmark resolution.
Want an SI influencer working for your brand?
Pick one from our roster and go live within 24 hours, or commission a character built only for you
You can check this yourself, for free
Verifying any of this costs nothing to begin. Quantum circuits can be built and run in a browser with nothing to install, so the claims in this article can be checked directly. There is a free online quantum circuit simulator that runs in your browser, which is enough to see how few values a small circuit actually emits and why spreading an image across many of them is necessary.
Beyond the free tier, simulation runs at a fraction of a cent per job, and time on real quantum processors is rented by the shot, in the region of $0.30 per task plus fractions of a cent per shot. An afternoon of genuine curiosity costs less than a coffee, which is a better basis for a strategy view than any vendor presentation.
Where quantum might matter to marketing first
If quantum computing reaches marketing, image generation is unlikely to be the first place. Areas built on optimisation and simulation, such as logistics, scheduling and some kinds of portfolio and pricing problems, are more frequently discussed as early candidates. For creative work, classical AI is moving quickly, and that is where the practical gains for brands are today.
Summary
Quantum computers generate images today. They generate handwritten digits at 28 by 28 pixels in greyscale, use a classical neural network to do the judging and split the picture into patches because the processors cannot represent a whole one. The distance to a photorealistic AI influencer is about four orders of magnitude in raw values, and the obstacles that matter most are generalisation, an analytical fidelity ceiling and gradients that vanish as circuits grow. It is a fascinating field and worth following, and campaign imagery today comes from classical generative AI.
Proklisi builds websites and apps, runs SEO, ASO and social media, and creates SI influencers for brands. Everything here comes from work we do for clients every week. About us or talk to the team