About GymDreams

Gay artist dreaming with AI — exploring male forms and representations, with Stable Diffusion, Midjourney, Photoshop AI, Python… Yale educated. Bauhaus philosophy. Works in design.

I am also the Founder of gayai.org, a non-profit organization that promotes knowledge sharing, and a community for male-focused AI artists.

On social media, I am known as GymDreams8, GayDefiance, or GD8.


Many of the works you’ll find on my feed delve into my personal experiences as a gay man. They weave together elements of autobiography, metaphor, and fiction.

Expressing the true essence of these works, or why I create them, is a challenge due to the intangible nature of the emotions they encompass. These feelings defy simple verbalization.

Greek mythology has held a profound influence on me since childhood, both as stories and as artworks. Even as a young child, I struggled to articulate precisely why I was drawn to them and why they continue to inspire the works I create.

However, what I can say is that by sharing and examining these pieces, I have gradually come closer to understanding myself. They serve as a form of reconciliation, bridging the gap between the intricate emotions I experience and my inability to fully elucidate them through logical reasoning.

Browse my works by tags in these categories:

  • AI Engine. The core rendering engine used to generate the image.
  • Artists. Source of inspiration and style.
  • Technology. Tools, models, methods, and techniques used.
  • Style and Medium. Visual vocabulary and medium used to describe traditional art mediums.
  • Subject. Subject matter depicted.
  • People. Name of an actual person if one is used.

More info


I am available for commissions. Please see Commission for more information.


In my professional design work, I utilize AI tools like Midjourney and Stable Diffusion to enhance my creative process. However, I have discovered that these tools have significantly improved my skills when I utilize them for personal projects, which I share under the name GymDreams8.

Typically, I begin with a concept or theme, forming a visual representation in my mind. However, during the initial exploration stage, I strive not to restrict myself too much, as the AI engines often surprise me with ideas that I hadn’t considered. I view these machines as partners, not only in generating images but also in brainstorming ideas. They excel in this area.

The next step involves multiple iterations of prompt manipulation to achieve the desired composition and styles. Sometimes, I incorporate additional images as input, as text descriptions can be limiting. The entire process involves a significant amount of art direction, along with post-processing using Lightroom and Photoshop to add the final touches.

Many of my works revolve around gay themes, encompassing concepts and ideas that have lingered in my mind for a lifetime, waiting to be unleashed. I sincerely hope that you will enjoy them.



  • Beginner-friendly AI generation tool.
  • Runs as a Discord bot.
  • Private source.
  • Midjourney.com

Stable Diffusion

  • Gold-standard AI generation tool.
  • Very powerful.
  • Very steep learning curve.
  • Open source.

Stable Diffusion UIs

Stable Diffusion is an open source technology. You can run it programmatically, but is is usually run with a GUI (graphical user interface). Here are some of the most popular UI implementations:

  • Automatic1111 SD WebUI
    • One of the most popular UIs out there.
    • Lots of plugins to extend features.
    • SDXL-capable.
  • SD.Next
    • Originally a fork of Automatic1111.
    • Now has a lot of unique features.
    • SDXL-capable.
  • ComfyUI
    • Node-based UI.
    • Powerful, but has some learning curve.
    • Relatively hard to pick up if you’re a beginner.
    • SDXL-capable.
    • One of the best UIs available if you want to use SDXL, the next-gen SD model.

Stable Diffusion as a Service

If you don’t have a power GPU (graphic card), you can use one of these cloud providers:

  • Mage.Space
    • One of the best cloud based SD providers. Highly recommended.
    • Beginner friendly
    • SDXL-capable

Stable Diffusion on M1/M2

Setting up Stable Diffusion on Apple Silicon (M1/M2) is not hard, but some technical knowledge is required. I wrote a guide on how to do it on my Docs site using Automatic1111.

Stable Diffusion in the Cloud

If you don’t own a powerful machine, you can rent GPUs in the cloud, then run Automatic1111 SD WebUI on them, with a docker image. This is what I use to run large jobs with Stable Diffusion.

I wrote a guide on how to do it with Vast.AI, a GPU rental provider, on my Docs site.

Photoshop AI

Introduced in Photoshop Beta since the summer of 2023, the AI engine is now fully realized in the live version.

  • It is terrible at generating images, but it is very good at fixing them.
  • I wrote a blog post about using it for out-painting, i.e. using it to extend the canvas of an image.
  • It is exceptional for inpainting as well. Images tagged with photoshop-ai are inpainted with Photoshop AI. I rarely use it for anything else.
    • Use it to fix problematic hands, eyes, feet, etc.
    • Sometimes an empty prompt works better than a text prompt.

Diffusers Library

If you’re fluent in Python, you can use the diffusers library to create your own AI generation tool. It is published by Hugging Face. Read the docs


Replicate is a machine-learning product targeted towards developers. For end users, It has a productivity-driven UI interface that allows anyone to create AI-generated images, as well as play with various AI models for LLMs, audio, video, etc. For the developers, it has a powerful API that allows you to create your own AI-powered apps. You can additionally use it to run your own models, and host them publicly or privately. Pricing is based on inference time.

I use this often when I want to make images quickly without setting up a full environment in the cloud, and when I need all the resources on my local machine for something else.


Before deciding on a name for a brand / product / identity, I always do a few sketches and make sure that I can design a decent logo for it before committing on the name.

For this identity, after a few sketches, I saw that I could make a nice mirrored form for the G and D, and that’s why I went ahead and registered the username GymDreams on DeviantArt — the original place where I started posting my AI works. There’s a longer backstory of where this name came from, but I’ll write about it in the future.

Once that account got some traction, I realized that I should probably start making profiles on other social media, but it didn’t take long for me to realize that this name faces some challenges since it’s often taken. I don’t want to randomly add spaces and dashes to the name — seeing that the logo I designed can be seen as a rotated figure 8, I decided to register @GymDreams8 on Instagram, Twitter, YouTube, and also bought the domain GymDreams8.com

Regarding the design, my art education was deeply rooted in the Bauhaus philosophy, so I also prefer forms that are made out of the basic shapes — squares, circles, and triangles. As such, logos that I design tend to have structures that can fit nicely in a grid, with proportions that can be described in x, 2x, 3x, 5x, etc.

The form of each letter is based on a circle with 5x radius or 10x diameter. They are separated by a gap that’s x long. Together, the two letterforms create a mark that’s exactly 21:10.


Although I identify as a designer, technology has been an integral part of my creative process for many years. I began programming at a young age, and my professional design work often lies at the intersection of art and technology. Creating AI art feels like a natural fit for me, as it allows me to leverage both my design and technical skills.

Regarding the construction of this website, I frequently use personal projects as an opportunity to explore new technical frameworks that pique my interest. This site was built from scratch using Nuxt 3, a web framework based on Vue 3. The content management system driving the site is Nuxt Content, an official module developed by the Nuxt team. This module enables me to write content in Markdown and publish it quickly.

For the CSS framework, I rely on Tailwind CSS. Tailwind is a utility-first CSS framework that empowers me to efficiently design new features and functionalities directly within the code. The site’s icons are provided by Font Awesome Pro, an excellent icon library offering a wide range of up-to-date icons. Additionally, Font Awesome's SVG icon feature allows for efficient bundling to reduce file sizes.

The images are hosted on Amazon S3 and optimized to various sizes using Cloudflare Image Optimization. Cloudflare’s content delivery network (CDN) caches the images, ensuring fast loading times for visitors. By utilizing direct repository integration with Github, the site is automatically built and deployed on Cloudflare Pages whenever I push a new commit to the master branch. Leveraging Cloudflare, with its extensive global network of data centers, ensures rapid loading speeds for visitors worldwide, regardless of their proximity to an origin server.


You can find me on my social media:

Feel free to reach out to me through direct messages on social media or by sending an email to hello@gymdreams8.com