How to Use Tess

This article is written for people who already have access to Tess, the AI image generator trained in a single visual style. If you don't have access, join the movement by coming to Tess and Signing In to create an account.

Like other software platforms, Tess is a tool that empowers creatives to make art. In this article, I'll explain how you can leverage Tess to generate illustrations for your website, newsletter, article, presentation, or project.

Get Prompt Ideas

First, you need to decide what kind of image you want to generate. Often, artists will browse through Pinterest or through art galleries to come up with concepts, assemble mood boards, and get inspired. Tess includes a couple of embedded tools to generate ideas for prompts:

• Idea to prompt: Enter an abstract idea, like "Late stage capitalism," to get ideas for visual prompts that could be used on Tess to make art relevant to that topic.
• Article to prompt: Paste the full text of an article to get ideas for art that illustrates the article. Great for making social media graphics to accompany blog posts or newsletters.

Tess uses AI to describe potential prompts and visual metaphors to accompany an article or visualize an idea.

To get inspired, you can also browse through the featured images on the Tess homepage or the Showcase on each model to see what other people have created and what prompts they used.

Craft a Good Prompt

AI is difficult to control, but as you use Tess you will learn more about what text inputs ("prompts") produce good results. The best prompts will visually describe the details of the scene, including what is in the foreground and background. Prompts can include multiple sentences. To create a usable image, you'll need to describe the layout of the image that you want.

• Use specific, literal descriptions. If you have an image in mind, describe it specifically. The prompt "a dog" will produce a less interesting image than "a golden retriever sitting on a beach at sunset."
• Describe the position of elements: Add variety to your generations by experimenting with composition description. What's in the background? The foreground? Some phrases to try: Zoomed in or Zoomed out; close-up; a scene; a simple icon.
• Employ artistic language to describe color and composition: Vibrant, unsaturated, impressionist, blocky, hard outlines, blurred bokeh effect.
• Include emotive words that convey the mood: If you're not sure exactly what you want in the picture, try to identify the tone of the output you want and communicate it through your prompt language. Mysterious, mystical, whimsical, ominous; bright, cheerful, laughing, joyful; serene; gentle; spooky; subdued.

The "Ideas" tool can help you elaborate and add detail to the description. Use it to get ideas for better image prompts. You also could upload a sketch to communicate what you want, then use the Image uploader on Tess to transform that image in your visual style.

Configure your generation settings

In the generations modal, choose the aspect ratio of your output. Tess supports vertical (9:16), landscape (16:9), square (1:), 3:2, and OpenGraph image sizes.

Tess users can also configure the strength of the model in the generated image. Increase the strength to make your images more similar to the artist's visual style or decrease to make it more generic. If an image is unrealistic, it may help to decrease the style strength to make the subjects more realistic and recognizable. We recommend starting with the default - 80% - and experimenting with different settings.

Make the image more or less busy (many details and textures vs less) using the "Low" and "High" Complexity toggle.

Video edited on Kapwing

Uploading an Image

On Tess, you can choose to start with an image and cast that image into the style of the artist. In the "Image" tab, upload an image and write a description of what items are in the image.

Your description has a strong influence on the output, so be sure to include details that you want to preserve in the output. Tess generates a caption for your image on upload, but we recommend editing and elaborating to improve your results.

Image to image generations do take a bit longer than text to image generations, so be patient! A thumbnail of the reference image will be included in the resulting generations.

Make Tweaks

Sometimes, a Tess image looks perfect – except for a small flaw. Maybe someone has an extra finger or leg, or maybe there are two figures when you asked for one. Tess has a built-in magic eraser in the "Remove Object" feature. Select the area that you want to remove, and it will blur the background over the selected area, making the object disappear.

You can also use "Faceswap" to sub in the face, making a face more realistic.

Use the "Generate Variations" brush to reimagine a section of the image. You might want to try out a new outfit, background, or expression for the image.

"Static Upscale" on Tess will increase the resolution of the image and fix minor precision issues in the image. The actual resolution of the image is written above the download button.

Download and Share

Once your Tess image looks good, it's time to download. All images generated on Tess have a subtle watermark overlay that is removed when you click "Download Image."

If you want to, you can contact the artist by clicking on their linked artist page. Artists often add their contact info and portfolio website, so you can reach out to them for custom commissions and projects if you like their style. Tess is designed to be a tool that enhances collaboration between artists and publications.

Tess is best used as a collaborator with the designer rather than the final output. Pull a downloaded image into Photoshop to make additional modifications or adjustments. Make sure to attribute the name of the model if you publish the image online.

Coming soon: Remove background, text tools, and more. Let us know how we could make Tess work better for you by emailing us feature requests and bugs at hello@tess.design.