Beyond the Frame: Workshop Instructions
Make your own images using artworks from Tate Britain’s collection and AI imaging software, Clipdrop. No prior knowledge of, or experience with, AI is necessary.
In the spirit of Queer and Now, you might want to consider issues of identity and representation as you create your new images. How might (or might not) AI tools enable you to generate images that better reflect and represent you and your identity or culture?
Want to see your work displayed? Just download it once you’re finished to see it projected here in the Studio as part of a revolving exhibition of images produced in this workshop. Please let us know if you don’t want your image displayed.
You can also use the paper printouts of the same Tate artworks which you can colour, draw on, and remake in the Studio.
If you have any questions or need help, please ask one of the workshop facilitators – Tate Collective Producers and students from Winchester School of Art. We will also be giving live demonstrations of the AI tools throughout the day.
A note on images: We are committed to ensuring that Tate Britain is a safe, inclusive, and respectful place for our staff and our visitors. Therefore, racist, homophobic, biphobic, transphobic, xenophobic, sexist, or otherwise offensive or harmful imagery, themes or language are not tolerated. Any images created in this workshop that feature such imagery, themes, or language will not be displayed and the individual(s) who created them may be asked to leave the workshop or the gallery by a member of Tate staff. Please be kind with your images.
Tutorial
You will be using Clipdrop, an image editing platform based on AI, to edit your chosen artwork. You can access Clipdrop through the Chrome web browser on one of the computers in the room.
You will find a selection of artworks from Tate’s collection in the Tate Britain Images folder on the computer. You can select any of these to work with.
In this workshop, you will be using a set of Clipdrop tools to edit your chosen artwork. These tools are: ‘Uncrop’ and ‘Replace background’ though Clipdrop has other AI imaging tools.
The following pages guide you through how to use these tools. Any question, just ask one of the workshop facilitators.
Using the tools
You can use one tool to edit your chosen artwork, or multiple tools. However, tools must be used one at a time before moving onto the next. The order in which you use the tools will impact the image you produce; it is up to you which order you use the tools.
After using one tool to edit your image, if you would like to use another tool to continue editing then you will need to download your edited image and then upload it into the next tool. Once you are finished editing completely, you will download your final image.
‘Replace background’ tool
This tool allows you to use text prompts to fill in, or edit, the background of an image.
(1) From the top bar menu under ‘Tools’, select ‘Replace background’
(2) Select an image from the Tate Britain Images folder and drag it into the box ‘Click, paste, or drop a file here to start.’
(3) The background – or what AI thinks is background – of the image will be removed. You can now use text prompts to replace the background with a new one. You can use the suggested prompt, generate a random prompt, or write your own. See our guidance below for more info on prompts.
(4) Clipdrop will generate images based on your prompt. You can click the ‘refresh’ icon to generate new images. Once you are happy with an image, click ‘Download’ to save it as your final image, or to continue editing with another tool.
‘Uncrop’ tool
This tool allows you to ‘uncrop’ an image – i.e. to extend beyond the frame of the image to imagine what might be there. The tool was launched just days before the Tate event. To find out more read Clipdrop’s blogpost. You can apply the tool an infinite number of times, to keep extending the frame. There are no text prompts.
(1) From the top bar menu under ‘Tools’, select ‘Uncrop’.
(2) Select an image from the Tate Britain Images folder (or the Downloads folder if you’re using an image you’ve started editing with another tool) and drag it into the box ‘Click, paste, or drop a file here to start.’
(3) Resize the image, ensuring it’s smaller than the outside frame, marked in blue, (the smaller the image, the more space will be filled in with imagined image). Then hit ‘next’.
(4) Once the image has been uploaded, Clipdrop will generate several new images. You can click the ‘refresh’ icon on any of these to ask it to generate a new one. Once you are happy with an image, click the ‘Download’ button to save it as your final image, or to continue editing with another tool.
Cleanup Tool
This tool allows you to erase selected parts of an image.
(1) Scroll down in the top menu under Tools to find Cleanup and select. Select an image from the desktop folder and just drop it in the box that says ‘Click, paste, or drop a file here to start.’
Use the ‘brush tool’ to paint over the areas you want to be removed. The size of the brush can be adjusted and the brush tool leaves a green mark on the area selected. When you have finished selecting the area press ‘clean’.
You will see the content removed and more of the background appear.
To find out more and to experiment with the full range of tools, visit Clipdrop →
Guides on Prompting
If you’re new to AI art, prompting or ‘prompt engineering’ may seem daunting. At its most basic, however, it’s just using words to paint a picture of what you want to create. A ‘prompt’ tells the AI model what we want it to generate. The more descriptive the prompt, the better the results.
A prompt is usually structed like this: {object} + {in the style of}
Examples:
“Manhattan at night in the style of Van Gogh”
“An avocado couch in the style of anime”
Core Prompt: A simple description of your central theme, subject, or figure (for example, rainbow, a dancer with a flag, skeleton etc). This is the basis of your prompt, but good prompts require more specificity. This is when we use style….
Style: Style is a crucial part of the prompt. Examples include: ‘impressionism’, ‘anime’, or ‘like the Simpsons’. If the AI model fails to recognise the style in the prompt, it will usually default to a common trope. For example, given the prompt of “landscape” + a style it doesn’t recognise, the model might generate a landscape in the style of an oil painting.
Artists Names: To make your style more specific, you can use artists’ names in your prompt. For instance, you can add “in the style of Pablo Picasso” or just simply, “Picasso”.
Finishing Touches: These are the additional elements that really help to make the generate image look like how you envision it. You can add anything you want here, but below are some examples that work well:
Highly detailed, surrealism, triadic color scheme, smooth, sharp focus, matte, elegant, illustration, digital paint, dark, gloomy, washed-out colors, sharp, dramatic lighting, beautiful, ambient lighting, epic composition etc.
In the spirit of the workshop and Queer and Now, you might want to consider issues of identity and representation as you write your prompts. How might you use prompts to generate new images that represent and celebrate your identity or culture?
We all want to feel seen and to express ourselves creatively – maybe with a little practice, prompt engineering can enable us to create images to express who we are. Or maybe AI isn’t quite there yet; let us know what you think!
Supported by:
