Mystic is the latest artificial intelligence image generation model on the market and was developed by the team behind AI image platform Freepik and its recent acquisition, Magnific AI.
The Flux-based model can create hyperrealistic images of people, places, and even obscure scenes, accurately follow prompts, and generate realistic text on images. Javi Lopez, creator of Mystic says it is the "most advanced AI generator to date with outputs directly in full HD."
To put Mystic to the test I came up with a mixture of complex, and very simple prompts to see how the AI would handle the requests. The focus is on photorealism but with a degree of weirdness, such as cats playing piano or a floating island with a waterfall.
Mystic is currently only available on Freepik and for premium subscribers but is likely to get a wider rollout in the future. Even without Mystic, it is one of the best platforms for AI image generation with a range of models, features and tools.
What is Freepik Mystic?
We're 40,000 people. You did it 🥳 We promised you something, didn't we? Enjoy Freepik Mystic for 24h. Use CODE: 40KFreepikLet's go! 🚀 pic.twitter.com/MI8wFVpGNRAugust 29, 2024
Mystic is a new AI image model by Freepik and Magnific that is a series of fine-tunes of Stable Diffusion, Flux and Magnific-built models. It was developed with input and image curation by a team of photographers, digital artists, VFX specialists and designers.
A Freepik spokesperson told Tom’s Guide its training included “many hours of in-house specialists with a very good eye tweaking every little detail.”
It is at its core a new workflow that activates the Flux mode, easily one of the best AI image foundation models currently available. In the workflow it triggers some Stable Diffusion processes, the fine-tunes made by the team and upscaling from Magnific.
“This combination delivers breath-taking visuals in high resolution (2k), without extra steps needed,” the spokesperson explained.
Creating the prompts
One area Freepik says Mystic excels is in prompt adherence, so I’ve created a series of very descriptive prompts with details such as the style and polish of a piano and facial expressions.
Freepik also offers prompt enhancement, so I’ve thrown in two very simple prompts to see how well it matches the idea I had in my head for that image.
I’ve shared all of the prompts below along with the output I got from running that prompt so you can try for yourself. These prompts will also work on a standard installation of Flux, Midjourney or Ideogram to varying degrees.
1. A Cat Playing a Piano
The prompt: “A cat, sitting upright on a grand piano, with its paws delicately pressing the keys. The setting is a luxurious, dimly lit room with velvet curtains and a chandelier overhead. The cat is a sleek, black feline with a shiny coat, and its eyes are focused intently on the keys as if it’s deeply engrossed in the music. The piano is polished to a high sheen, reflecting the soft glow of the chandelier, and there’s a faint sense of magic in the air.”
I think this one captured the prompt well. It followed the instructions for the cat and the piano but I think it could have anthropomorphized the cat a little more.
2. Person with a Floating Head
The prompt: “A young person standing in a minimalist room, with their head floating a few inches above their neck. The head is fully detached, yet it maintains perfect alignment with the body, as if nothing is out of the ordinary. The expression on their face is calm and composed. The room is stark and modern, with clean lines and soft lighting, enhancing the surreal nature of the scene.”
The scene looks great. It is incredibly realistic but it didn't detach the head. I re-ran the prompt multiple times and not once did the head appear above the neck.
3. Giant Turtle with a City on its Back
The prompt: “A colossal turtle slowly moving through a vast desert, with a thriving, bustling city built on its massive shell. The city has towering skyscrapers, suspension bridges, and lush green parks, all carefully balanced on the turtle's back. The turtle itself is ancient, with a weathered shell that tells stories of countless journeys. The sky above is a deep blue, with clouds casting shadows over the desert landscape. The scene captures the blend of nature and urban life in a fantastical, dreamlike setting.”
This one has a Terry Pratchett vibe and Mystic captured exactly what I was hoping for with the prompt. In fact, I think it did a better job as it also weathered the city.
4. Woman with Hair Made of Fire
The prompt: “A fierce and powerful woman with long hair made entirely of flames, standing on the edge of a cliff overlooking a stormy ocean. Her fiery hair crackles and dances in the wind, illuminating her face with a warm, glowing light. She’s dressed in a flowing, dark gown that contrasts with the vibrant flames. The sky is filled with dark clouds and flashes of lightning, and the waves crash violently against the rocks below, adding to the dramatic atmosphere.”
I love the contrast of the hair and the waves. What I think Mystic has done well here Is capture the lighting perfectly.
5. Dancer with Shimmering Silhouette
The prompt: “A graceful dancer mid-performance, her entire body outlined by a shimmering, ethereal glow. The background is a dark, empty stage, with a single spotlight focused on her, but it’s the glow that defines her form. She’s captured in a fluid, elegant pose, with the light shimmering like stars around her. The contrast between the darkness and the radiant silhouette creates a mesmerizing, almost otherworldly effect, highlighting the beauty and grace of the dancer.”
A simple, elegant scene. When I tried animating it with Runway it captured the 'glow and stars' concept, adding a swirl of stars moving with the arms and legs.
6. Robot painting a starry sky
This is the first of the two very basic prompts. The title is the entire prompt I used “robot painting a starry sky”.
I then asked ChatGPT to describe the full prompt likely used to generate the image: "A futuristic scene featuring a humanoid robot with a sleek yellow design, standing against an outdoor wall during sunset. The robot is intricately painting a glowing spiral galaxy on the wall, with the galaxy emitting a soft blue light that contrasts against the darkening sky. The setting is urban with blurred city lights in the background, emphasizing the advanced technology and artistic expression of the robot."
I actually love how Mystic interpreted the simple prompt. I don't think it is how I would have described it myself as I'd have it on its back painting the actual sky but this works really well.
7. Floating island with a waterfall
This is the second of the two very basic prompts and the final prompt. The prompt I used was “floating island with a waterfall”.
I then asked ChatGPT to describe the full prompt likely used to generate the image: "A surreal fantasy landscape at night featuring a massive floating island with lush greenery and cascading waterfalls flowing into a calm ocean below. The full moon dominates the sky, illuminating the scene with a soft, ethereal light. The island is suspended in mid-air, with the waterfalls creating a mystical atmosphere. The scene is peaceful and otherworldly, with the distant horizon barely visible under the moonlight."
This is exactly as I imagined it and Mystic captured the concept from a simple five word prompt perfectly.
Final thoughts
You can see the 'Flux' in the images generated by Mystic, especially if you spend a lot of time with Flux tools. But, its generations are significantly more realistic and you can see the work that went in to fine-tuning the model and in the workflow.
Mysic is a genuinely impressive model and its output animates surprisingly well with Runway or Luma Labs Dream Machine, turning the still image into a realistic AI-generated video shot.
Midjourney is still the industry leader for its broad set of tools, the massive sref library and its very realistic and stylistic images. However, companies like Freepik, Leonardo (now owned by Canva) and Ideogram are giving it a run for its money.
More from Tom's Guide
- Best food delivery services: Grubhub vs Uber Eats vs Doordash
- The best cast iron skillets 2024: Tested and rated
- How to make images with AI using Leonardo
