Beta Have fun experimenting and please report any issues. If you want to help us, please share your top results in public ❤️ Pro tip: Upscale your favorite transforms to really make them look awesome!
The selected file is not a valid image.
The image that transfers the style. For example, if it's an image with a cartoon style, you'll transfer that style to the input image.
The selected file is not a valid image.
You can guide the generation process and influence the style transfer with a descriptive prompt. For example, if the reference image is a Van Gogh painting, adding something like "An oil painting of a pretty woman in the style of Van Gogh" will be helpful.
You can also use your imagination to alter images: transforming a landscape from summer to snowy winter, changing a person's eye color, or just crazy things like adding huge green aliens from outer space attacking a city.
IMPORTANT: You can emphasize specific parts of the prompt by using a number in parentheses, ranging from 1 to 1.4, like "(beautiful green eyes:1.3)".
Allows you to go from your original image textures and colors (0%) to fully transferring the style of the reference image (100%), guided by your prompt.
Allows you to maintain the general structure of your image (lines, borders, framing, etc.). Combined with "Style strength", you can do things like fully transferring the style but not the structure: something cool that will be near to a completely new image but with the style of the reference image and your prompt.
Turn this on when you're working on a portrait.
• "Standard", "Pop", and "Super Pop" offer different vibes for portrait styles. Which one works best really depends on the case. Usually, "Pop" and "Super Pop" are cooler, and "Standard" is more neutral.
• The "Beautify" option will enhance the person's face, but it may result in a slightly less accurate resemblance.
Different ways of transfering the style:
• Faithful: best at transferring the style of the reference image. However, at high levels of "Structure strength", it can create somewhat overloaded images.
• GenZ: adds a more artistic, saturated, and colorful touch.
• Psychedelia: similar to GenZ but with softer tones and a dreamy vibe.
• Detaily: transfers the style a bit less, but can help gain sharpness.
• Clear: similar to "Detaily", but a bit softer.
• Donotstyle: almost disables the style transfer from the reference image, which is very useful when the goal is not to transmit the style of the reference image, but simply to modify the final image through your prompt and "Structure Strength".
• Donotstyle Sharp: similar to Donotstyle, but with more saturated colors and, in most cases, crisper lines and shapes.
Balanced: very well-balanced, beautiful, and useful in most situations. It can work for realistic images, but it mainly excels in digital drawing, design, and subtle details.
• Definio: definition, balanced, clean, detailed and realistic.
• Illusio: for illustrations, paintings, drawings, sketches and digital art.
• 3d Cartoon: perfect for 3d cartoons :)
• Colorful Anime: perfect for colorful drawings and anime. Try a prompt like "lots of flowers, sunny summer, vibrant colors, blooming" and marvel at the explosion of colors.
• Caricature: caricatures and colorful cartoons.
• Real: extremely detailed. Good for realistic images. Works well with "Donotstyle" flavors.
• Super Real: for realistic and detailed images. Tip: Use the "Clear" flavor for even more realistic and defined results.
• Softy: smooth and soft results. Tip: You can use it as a smooth base for future high-resolution upscaling.
When this option is enabled, using the same settings will consistently produce the same image. Fixed generations are ideal for fine-tuning, as it allows for incremental changes to parameters (such as the prompt) to see subtle variations in the output. When disabled, expect each generation to introduce a degree of randomness, leading to more diverse outcomes.
First, drop your input and reference images in the area above to transform.
The selected file is not a valid image.
You can guide the upscaling process with a descriptive prompt. If the image you're upscaling is AI-generated, reusing the original prompt here can greatly improve the upscaling quality!
You can also use your imagination to alter images: for example, transform a city into ruins during the upscaling process (typically after at least 4x upscale), change a person's eye color, or give a portrait the look of a famous person.
IMPORTANT: You can weigh specific parts of the prompt by using a number in parentheses, ranging from 1 to 1.4, like "(beautiful green eyes:1.3)". Magnific AI sometimes ages faces in its outputs. To counter this, adjust your prompt weights, for example: "(young woman:1.3), (cute young face:1.2), (cute:1.2)".
Allows the AI to "hallucinate" additional details, achieving greater realism at the cost of moving further away from the original image. Here's where Magnific's magic shines! But be careful: really high values can lead to some pretty strange results.
Increases definition and detail, though very high values can result in images with an artificial appearance or blotches.
(Advanced) Increasing this value will make the generation more closely resemble the original image, but very high values can result in blotches or a dirtier look. Lower values give more freedom to the generation at the cost of moving further away from the original image.
(Advanced) Control the strength of your prompt and intricacy per square pixel:
- Lower Fractality: less detail, but typically resulting in fewer glitches. If vertical bands appear in your image, reducing Fractality might resolve it.
- Higher Fractality: amplifies your prompt in increasingly smaller areas of your overall image. E.g., if your image is a rose and you use "A photograph of a rose" as your prompt with a high Fractality value, smaller rose-like details may emerge within the main rose. A bit crazy, but this can be useful sometimes for artistic purposes. For intricate images, like fantasy maps, high Fractality at resolutions up to 10k can generate astonishing details like rivers, mountains, and cities suggested in your prompt.
(Advanced)
• Illusio: better for illustrations, landscapes, and nature. The smoother one. Also good for the first pass of several upscales. Removes JPEG artifacts.
• Sharpy: better for realistic images like photographs. It provides the sharpest and most detailed images! However, it doesn't remove JPEG artifacts and can sometimes even create "fake JPEG artifacts".
•Sparkle: also good for realistic images. It's a middle ground between Illusio and Sharpy. Removes JPEG artifacts.
For multiple upscales, you can experiment: start, for example, with Illusio and finish with Sharpy to find the "sweet spot" for the style of your images.
First, upload an image to upscale using the drop area provided above.
×
×
I think this is the beginning of a beautiful friendship 🙂 Use the controls to upscale your first image!
