The rise of AI: Using AI to generate animation graphics
In this series of blogs, we’ll be exploring how AI could be used during different stages of our production process. In our last post we discussed the pros and cons of using AI to generate a video or animation script. In this post we’ll focus on using AI for the next step of the production process, to generate animation graphics such as illustrations and storyboards.
Firstly, let’s look at illustrations. Once a script has been written and signed-off, the next phase in our animation production process is the development of illustrations. This can be anything from characters and scenes to icons and infographics. We design bespoke illustrations based on the following elements:
- The script – what assets need to be created to visually reflect the content of the script?
- The audience – who is the audience and what kind of visuals will they find engaging?
- Branding – does the client have brand guidelines that we should adhere to?
- Representation – do the characters have specific demographics that need to be represented in the design?
- The content – if the animation deals with sensitive topics, if so, can the illustrations help soften the tone of the content?
To get each of the elements listed above right, we work closely with our clients to understand the subject matter, the objective of the animation and the target audience. We will source reference images from the client, or during our own research, and provide clients with sample illustrations for review. Once the illustration style has been approved, we then develop the rest of the animation illustrations that will feature in the animation.
Using AI to generate illustrations
There are now a variety of AI tools that can help with the creation of illustrations. These tools leverage the power of generative adversarial networks (GANs). GANs is a type of machine learning framework that enables users to almost instantly generate images based on text commands. Let’s look at some of these tools below:
DALL-E 2 from OpenAI was one of the first AI image generators to appear on the market. It can be used to generate illustrations from text prompts. After writing a prompt, you are provided with four AI-generated images. You can then either download these images, create more variations or refine your prompt. This tool is useful for users looking for inspiration or illustrations. You need to buy credits to use DALL-E 2.
Adobe has released the AI art generator Firefly. Firefly is free to use and offers a range of functions. The text to image tool enables you to provide a text prompt that provides you with four AI generated images. You can then customise elements such as the aspect ratio, content type (photo or art), style and strength, colour and tone, lighting, composition and you can even upload your own reference images so that it matches your preferred style. Other tools include Generative Fill where you can use a brush to remove objects, or paint in new ones from text descriptions, Text Effects where you can apply styles or textures to text with a text prompt, and Generative Recolour where you can generate colour variations of vector artwork from text descriptions. We used Firefly to generate the header image of this post using the prompt “A crowd of different characters in a hospital”.
Shutterstock assets are used by many companies that don’t have the in-house capabilities to create photos, music, illustrations and vectors. They’ve now released Shutterstock AI. Like the tools above, you can generate AI images using a text prompt. These images can then be refined by type, style and perspective. The tool is free to use if you have an account, and images can be downloaded and licenced to use in creative projects just like other Shutterstock assets.
The tools mentioned above open creative possibilities for companies. They can help us to generate new illustrations, generate illustrations based on existing assets, and adapt existing assets using the power of AI.
How AI generated illustrations compare to artist drawn illustrations
To compare the capabilities of AI tools to generate illustrations compared to illustrations drawn by an artist, we’ll look at one of our previous projects.
We were commissioned by NHS Health Education England to produce a set of nursing character illustrations that included an adult nurse, a midwife, a paediatric nurse and a Return to Practice nurse. The illustrations needed to be the same style as other NHS characters we had previously produced for NHS Health Education England. We worked closely with the client to ensure the characters were representative of the demographics of the nursing workforce. Below is the set of character illustrations that were produced:
Now let’s compare one of these illustrations to what we could generate using an AI tool. For this exercise we’ll aim to generate the midwife illustration using Adobe Firefly.
To give Adobe Firefly the best chance of generating an illustration that matches the clients brief and required style, we shared the description the client provided us with and uploaded a similar illustration in the NHS style. After some tweaks to the settings and the prompt, here are the illustrations that Adobe Firefly generated:
As you can see, the illustrations are close to the style of the reference image we shared, but they’re not the same. Rather than just using the reference image for the illustration style, Adobe Firefly kept the white lab coat that the character in the reference image was wearing despite it not being part of the prompt. It also had difficulties interpreting the sunflower lanyard element of the prompt. The character needed to wear a sunflower lanyard as it was a visual indicator for a hidden disability.
Now let’s look at how we can use AI tools to help generate animation storyboards.
Once the illustrations are in place, we’ll move onto the next design phase of a project, the animation storyboard. A storyboard is a static graphic representation of the animation that features key scenes from the script. We develop storyboards so that clients have visibility of what the final animation will look like before we start the animation process. This is also an opportunity for us to bring all the content assets together. Having a well-designed storyboard often makes the final animation phase much smoother.
Using AI to generate storyboards
In addition to using AI tools to generate animation illustrations, they can also be used to generate storyboards. One such tool is StoryboardHero that combines all the disciplines we have discussed so far.
StoryboardHero can be used to generate a script, images and a storyboard. All you need to do is enter the animation idea you have, and it then generates a script and divides it into dialogue and scene directions. These can be regenerated and edited until you are happy with the content. Once the script is ready you have the choice of using it to generate AI images, or you can buy credits and request the images to be drawn by an artist.
StoryboardHero isn’t the only tool that offers these capabilities, there is also Boords and Krock. With all these competing AI tools at our disposal, let’s look at the pros and cons of using AI to generate animation graphics.
As we’ve outlined above, our process for developing illustrations can take time. From discussing the project with the client to the time it takes to create and refine the illustrations. One of the biggest benefits of using AI to generate graphics is the time you can save.
A knock-on benefit of saving time is saving money. Using AI tools also means that you don’t need to invest in creatives, either in-house or through outsourcing. Cutting down production costs can enable smaller companies and freelancers to create visual and animated content.
AI tools can also be used by creatives looking for inspiration. If you’re starting a project and want to create a mood board to reflect different ideas or styles, tools such as DALL-E 2, Firefly and Shutterstock AI can be used to instantly generate image variations.
Lack of precision
Although AI is time and cost effective, it doesn’t always get it right the first, second or third time. Every image that’s generated is born from a text prompt, but even when you specify the exact image you want, you don’t always get the results you’re looking for. When working with an artist you can explain the changes you want to make, but with AI generated images you must either regenerate the image or change the wording of your text prompt.
Lack of creativity
Using AI to generate graphics can result in a lack of creativity and originality. This is because AI tools base the images they generate on existing images, data and algorithms. These tools can also learn by analysing other artists work without their permission which brings us onto the ethical concerns of using AI tools.
As we’ve mentioned above, AI tools learn from art that has been created by artists, but the artists may not have given permission for this to happen. An AI tool may generate art that’s similar to an artist’s existing work, without crediting them, notifying them or compensating them for it. The rise of AI tools can also lead to reduced work and revenue for artists.
As we’ve discussed, integrating AI into your animation production process to generate graphics can come with both advantages and disadvantages. Although AI image generation tools can save you time and money and boost creativity, it can leave you with generic content that lacks detail, originality, and in some cases, accuracy.
If you’re looking to develop illustrations for an animation or creative project, contact us today at email@example.com.