Tutorial #1: How to use Vision FX

In this tutorial, we’ll demonstrate how to use the AI Vision FX plugin. This plugin can be used with PaintShop Pro, CorelDRAW, Corel PHOTO-PAINT and Corel Painter, as well as other applications including Adobe Photoshop and Affinity. In this tutorial we’ll be using PaintShop Pro.

Launching Vision FX

Our first example is a photo of a forest, which has been resized to 512 pixels on the longer side. 512 is the preferred size for generative AI, and this size will optimize processing time. As you’ll see, generative AI is resource-heavy, and smaller images get processed much faster.

In order to preserve the original image, you can duplicate the background layer and leave the copied layer active. Right-click on the background layer in the Layers palette and select Duplicate.

From the main menu, choose Effects > Plugins > Vision FX. This will launch Vision FX in a new window.

• If you are using Corel PHOTO-PAINT or Corel Painter, the plugin will also be listed under the Effects menu.
• If you are using CorelDRAW, you will first need to have a raster image open, and the plugin will be listed under the Bitmaps menu.

Vision FX interface

Here are the main components of the Vision FX interface:

1. A sample prompt is provided at the top right, with comma-separated terms specifying a detailed digital painting in the style of Vincent van Gogh, with mystical colors and beautiful lighting.
2. The Strength slider controls how much the AI-generated results will deviate from the original. As strength increases, less of the original image will appear in the results.
3. Similarly, the Text Guidance slider controls how strictly the text prompt will be adhered to.
4. The Starting Seed is a random number that can be regenerated if needed. This is the number used to generate the first result, and the other four will use consecutive seed values starting from this number.
5. The Steps slider controls quality and speed. A high number of steps will produce better results but will take longer.
6. Click Run and Vision FX will start generating five AI images.
7. The generated options will appear in the Preview section at the bottom of the interface.
8. Click on one of the preview images to view it in the main window, and use the interactive slider to compare the AI result with the original image.

Generating AI images

Until you become familiar with how the individual settings affect your results, try starting with the default values for your first batch of AI generated images. When we use the default prompt terms and settings on this image of a forest, the trees in each of the preview images are more or less true to the original, while the AI engine has added the van Gogh style specified in the prompt. Some options even have additional van Gogh elements such as structures and people.

If you don’t like any of the results, you can click the Reset button, which fills in a new random Starting Seed value. Clicking Run again will produce another five options based on the same settings.

TIP: At any time, if you want to save your settings and seed values to use again in the future, you can make a note of them or take a screen capture.

When you are happy with a result, select it and click Save. This closes the Vision FX interface and brings the image back into PaintShop Pro (or whatever software application you launched the plugin from). In the Layers palette you can see that the duplicate layer has been replaced with the AI result you chose.

If you want an image that is larger than 512 pixels as we’re using, you can enlarge the AI generated image by using the Image > Resize command and the Advanced and AI Powered settings to upsample the image.

Adjusting settings and comparing results

If you want to see what results would be produced using different settings in Vision FX, you can work with multiple duplicates of the original image in the same PaintShop Pro file, and then compare results. To do this:

1. Select the original background layer in the Layers palette.
2. Right-click on the layer and select Duplicate.
3. Select the duplicate layer to make it the active layer, then launch the Vision FX plugin.
4. Make adjustments to your prompts and/or settings and generate a new batch of AI images.
5. Select one of the AI images and click Save to return to PaintShop Pro.
6. Use the Visibility (eye) icon beside the layer in the Layers palette to toggle layer visibility on and off to compare the different duplicate layers.

In our example, we have three duplicate layers of the original forest photo. For each duplicate, we used different Strength and Text Guidance settings to generate the AI images. As we increased these settings, the results looked less like the original forest image and more like a Van Gogh painting.

Creating and modifying text prompts

Experimenting with different prompts can be a lot of fun when working with AI image generating programs, but it takes some time to learn how to write a prompt that will result in the image that is in your imagination. But even the unexpected results can be delightful! Here are some tips for building a prompt.

1. Start with a clear statement describing the main subject of your image, for example “futuristic sci-fi city.” The order of prompt terms can make a difference in the results, so put the most important terms at the beginning of the prompt.
2. Include additional terms to describe other elements you’d like included, such as “alien invasion, skyscrapers under attack, spaceships.”
3. Describe the mood you’re envisioning, such as “ominous, dark, dramatic, and dystopian.”
4. Include the artistic style (e.g., photo-realistic, cartoon, comic book, oil painting, etc.) and if relevant, you can also include a particular artist or photographer.
5. You can even include terms to describe the lighting (e.g., “golden hour”), the color scheme, and the level of detail.

The best practice is to be ultra-specific with your prompts and start with a few terms that summarize the end result you’re looking for. Subsequent terms should mainly contain adjectives that describe or enhance the style description.

It’s important to remember that building prompts to generate AI art is an iterative process. Don’t hesitate to refine your prompt if the initial results don’t match your expectations. Experiment with different phrasings or add more details to guide the AI better.

Creating characters

Now let’s look at what we can do with a portrait. In this example, we’re starting with a photo of a woman on a beach that we want to transform into a character.

We’re using a set of prompts for a sci-fi, high tech comic book style and also including popular rendering terms such as photorealistic, volumetric lighting, octane render, mix of western and eastern art styles, detailed 4k.

For another example, we transformed a photo of a young woman outdoors…

… into a manga style portrait image.

You can even bring your pets into the fun. Here’s a photo of my cat in a tie…

… turned into a sheriff in a dusty old-time saloon.

Tutorial #2: 10 Tips for better AI art

With the Vision FX plugin you can create AI art from a photo, an image, or even a blank canvas. But creating AI generative art is not a one-click process or a quick and immediate effect, such as a PaintShop Pro script or Photoshop action. Instead, think of Vision FX as a powerful tool in your artistic toolkit. It takes a bit of practice, but the plugin can help bring your ideas and imagination to life.

In this video, we share 10 tips to help you get the most out of Vision FX.

Tip #1: Be descriptive

While there isn’t a specific number of keywords that work best for any image, using several strong and descriptive keywords in your prompts helps Vision FX produce more intricate and detailed results.

In this example, we’re starting with a photo of a camper van. Using just two keywords – “spaceship” and “science fiction” – and keeping the other settings at their defaults, gives us a decent starting point.

Starting off with a short list of keywords is a good way to quickly generate some rough ideas that you can then build on. In this example, we’ll increase the Text Guidance scale, and expand on the prompt terms. It’s important to be specific and detailed – the more details you include, the more likely you are to get the results you envisioned.

When the prompt is expanded to include some additional terms such as “aliens, dystopian, battles, lasers, explosions, scary, dramatic, intricate, and detailed,” the resulting images have more of these details and elements in the spaceship and the landscape.

Tip #2: Reduce the number of generated images for faster processing

When you’re experimenting with different settings and/or prompt terms, it can be time-consuming to generate a whole full set of five AI images each time. To try out a new idea while saving some time, you can generate a partial set of new AI images instead. To do this, select the fourth image, adjust the prompts or settings, and click Run to regenerate only images 4 and 5.

In this example, the prompt was updated to include “in the style of Andy Warhol” and the Strength and Text Guidance settings were increased. We selected the fourth image from our previous results and clicked Run to generate 2 new images. This allows us to see what kind of results the modified prompt and settings will produce, but the processing time is much faster since we are only generating 2 new images.

Once you’re happy with the results produced from the modified prompt and settings, you can use them to produce a full set of 5 results:

• Select one of the new images.
• Copy the prompt and make note of the settings.
• Select the first result and apply the new prompt and settings.
• Click Run to generate 5 new images.

Tip #3: Lower the Steps setting for faster processing

Another way to speed up the processing time when you are experimenting with different prompts and settings is to keep the Steps setting at the minimum. The Steps setting controls how the image is processed or diffused, which affects the quality of the result and the amount of processing time required.

In this example, these images were generated with the default van Gogh prompts and settings. Using the minimum Steps value of 20 produces relatively quick results, but not of the highest quality. These initial low-step images have loose detail and some artifacts.

Although these images are of lower quality, they are good enough to see what kind of results the prompt and settings generate. If you’re happy with the style of results, you can then increase the Steps setting and generate a new set of images. When the Steps setting was increased to the maximum (75), the resulting images took longer to process but have smoother lines, higher detail, and no artifacts.

Tip #4: Use a screen capture or snipping tool to save settings

If you like a particular set of results, you’ll only get the same results again if all settings are exactly the same. This means you’ll need to use the exact same prompt terms in the same order, the same Strength and Text Guidance values, and the same Starting Seed number. The quickest way to save these settings is to use the Windows snipping tool, or another screen capture tool, to take a snapshot of the Vision FX window. Save the captured image for future reference when running the plugin again.

Tip #5: Use upsampling to create larger images

Smaller images get processed faster by Vision FX. 512 pixels is the optimal starter size and is also the default size produced by Vision FX. But if you want your final image to be larger, try upsampling your AI results in your host software.

In this example, we selected one of the spaceship results and clicked the Save button to bring the AI image back into our host software, which is PaintShop Pro.

To upsample in PaintShop Pro, go to Image > Resize. In the Resize window, choose a resize method and enter the new size (we’re using Resize by Percentage and increasing by a percentage of 400). The Resize window also has Advanced and AI-Powered Settings, which you can enable if you want to choose the resampling method and settings (we’re using AI-Powered, in Illustration mode).

After clicking OK, the image is now 4 times larger with resolution similar to the original.

If you are using the Vision FX plugin in CorelDRAW, go to Bitmaps > Resample to upsample your AI image. If you’re using Corel PHOTO-PAINT, go to Image > Resample.

To learn more, watch the full tutorials on upsampling in PaintShop Pro or upsampling in CorelDRAW.

Tip #6: Create a landscape image from scratch

To generate an interesting landscape image, you don’t have to start with a landscape photo. You can start with a completely blank image, or colored background. In this example we’re starting with a PaintShop Pro image that has been filled with the Dusty Skies gradient.

In Vision FX, we’ll start with the prompt terms “mountains, Autumn leaves, lake” and see what kind of results we get.

The Strength and Text Guidance sliders need to be at the maximum value since there’s nothing else in the image to go on, and we’ll set the Steps value low for quicker generation.

After clicking Run, we can see that in all generated images, the gradient is nicely incorporated in the background.

These first images are a good start, but we want the results to include a lake. The order of prompt terms can make a difference, so we’ll rearrange the prompt terms so that “lake” is the first term. After regenerating a set of images, we have a few lakes, but the results are still pretty basic.

When relying on text guidance alone, it’s very important to have a healthy number of prompt words and phrases.

As we’ve seen, just a couple of prompts means you have less control, which can be great for starter ideas. When you’re ready to build on your prompts, more is better, so long as they make sense together. Think of the image you are trying to evoke, and the ways you would describe it. Remember, this AI model is built from descriptions of existing images and will use those descriptions to attempt to reproduce what you’re requesting.

Now we’ve added a few more terms like “colorful boats in the lake, Impressionist style, and relaxing and peaceful atmosphere” and bumped up the Steps for higher quality.

Each landscape now is now much more detailed and tailored to our prompts. Even so, not every option shows all of the elements we want, which means that some trial and error is usually needed for adding, changing, and re-ordering prompts.

Tip #7: Create a character from scratch

As with an AI-generated landscape, you can also use Vision FX to generate a new character. In this example, we’re starting with a completely blank image, and entering prompts to generate a ballet-dancing cat.

As before, we’re using maximum Strength and Text Guidance. This is a great way to generate from-scratch ideas for characters, book covers, posters, etc.

Tip #8: Refine your AI art in your host software

You can combine the tools of your host software with the power of Vision FX to craft exactly the image you want. In this example, we’ve used Vision FX to generate a few ancient Chinese parchment versions of a photo of a cardinal.

We’ll click Save to bring one of the images back into PaintShop Pro, which replaces the original photo in the layer we duplicated for this purpose.

TIP: before bringing your image into Vision FX, it’s a good idea to preserve your original image by creating a duplicate layer to use in Vision FX.

Next, we’ll duplicate the AI layer to preserve the original Vision FX image, then open the Instant Effects palette, and apply the Aged Texture effect from the AI-powered category.

With this layer active, that has both AI and PaintShop Pro effects, we can go back to Vision FX and apply some new prompts to add a watercolor on parchment effect.

This is just a simple example of how you can use Vision FX in conjunction with your host software. Any of the host software programs that are compatible with Vision FX have a vast array of tools, adjustments, and effects that you can use as part of your photo editing, graphic design, or digital art workflow.

Tip #9: Fix AI portraits with your host software

One area that AI in general doesn’t always get right is facial features and anatomy. But you can use tools in your host software to fix problem areas.

In this example we’ve generated some Renoir-style results from a portrait photo, and in each one, the face is a bit off. We’ll choose one to work with, and click Save to bring it back into PaintShop Pro.

Because we used a duplicate layer of the original photo to work with in Vision FX, we still have the original photo on the layer below. There are a few ways we can touch up the face.

Touch-up method #1: The Eraser tool

One way to touch up the portrait is with the Eraser tool. Select the tool on the toolbar and then on the Tool Options palette you can adjust the size and other brush settings. For our image, we’ll use a small brush size with a low hardness and reduced opacity. The low hardness and opacity will give our brush strokes soft edges and only partial transparency, which will help blend the two layers. Sweeping over problem areas on the AI image allows the original photo on the layer below to show through. You can use the visibility (eye) icon to toggle the layer on and off to compare the photo with the AI image.

To learn more about the Eraser tool, watch the full tutorial.

 

Touch-up method #2: Layer blend modes

Adjusting layer blend modes can also produce nice effects, such as Screen in our example. To try out different blend modes, just select one from the dropdown list at the top of the Layers palette.

Touch-up method #3: The Clone Brush tool

Another PaintShop Pro tool that can help with facial features is the Clone Brush tool. In the Layers palette, we can turn off the visibility of the AI layer and select the original layer, then right-click on the right eye to clone it. Back in the Layers palette, we’ll turn on the visibility of the AI layer and select it, then click to replace the eye.

To learn more about the Clone Brush tool, watch the full tutorial.

Tip #10: Convert AI images to vector images

With vector-based software such as CorelDRAW, you can easily convert a Vision FX generated image into a vector file. The advantage of converting a bitmap (also called “raster”) image such as a JPG or PNG to a vector image is that vectors are fully editable and can be scaled up to large sizes without a loss of quality.

In this example, we used Vision FX to generate an elephant logo from a blank image, using prompt terms including “logo, corporate, elephant, flat image, clean lines, symmetrical, and black and white.”

After saving the image as a JPG, we imported it into CorelDRAW. To convert it to a vector image we’ll use the PowerTRACE feature. There are a few ways to open PowerTRACE:

• Right-click on a bitmap to open the context menu and choose a trace option.
• Select the image and choose a trace option from the Trace Bitmap dropdown menu on the property bar.
• Under the Bitmaps menu, select a trace option.

For this image, we are using the Outline Trace > Detailed Logo option. In the PowerTRACE window, you can adjust the settings for detail, smoothing etc., then click OK to start the trace.

After the trace is completed the new vector image will be placed on top of the original bitmap image in the drawing window. In the Objects docker, we can see the group of curves that make up the vector. Because this vector design is fully editable, we can reshape the curves, change the colors, resize, add text or effects, etc. and use this design for T-shirts, posters, marketing materials, and all sorts of graphic design projects.

To learn more about the PowerTRACE feature, watch the full tutorial.

We hope these tips help you as you explore the possibilities of what you can create with AI generative art.

Learn more about Vision FX