Adobe InDesign is the most powerful page layout tool in existence. It is also extremely well-exposed to automation, and is available in server form. It is thus well-suited to leveraging Artificial Intelligence (AI) in the creation of documents and images. Olav Martin Kvern and I recently led this discussion with the Los Angeles InDesign User Group (which you’ll note is not so LA-centric any more):
Below is a text version of the content we prepared; a brief survey of how designers and content creators can extend this incredible product using the latest advances in AI technology.
Adobe InDesign Extensibility
One reason that AI and InDesign are a natural combination is due to InDesign’s incredible extensibility. As I explained elsewhere, InDesign is the most extensible page layout tool in existence, due to its architecture, as well as Adobe’s commitment to this capability. It can be extended with C++ plugins, ExtendScript, CC Extensions, and the more recent UXP technology from Adobe. The server-based version of the product, Adobe InDesign Server, can be hosted in the cloud, so it is perfectly possible to connect InDesign, at scale, to AI applications via web services. InDesign’s robust extensibility provides a very solid foundation for combining AI with InDesign in any number of creative ways.
Human Intelligence: Old School InDesign Automation
There is a fine line between “rules-based automation” and Artificial Intelligence. Many approaches to InDesign automation in the past looked very much like artificial intelligence, yet technically probably were not. Ever since InDesign came on the scene, companies like Silicon Publishing have used its incredible automation capability to create document solutions that could be considered “intelligent.” For example, we automated the creation of thousands of cruise booklets, each one unique in terms of the content, the layout, and the pagination. The booklets look hand-crafted, although the process is entirely lights-out. Advertisements are intelligently inserted into the pages of the document, filling available whitespace with targeted content for the individual recipient.
It’s an intelligent flavor of automation, but all the logic was defined by humans who knew page layout, typography, and targeted messaging. AI can enhance this sort of solution, with automation based on machine learning. Artificial Intelligence might be applied, for example, to generate some of the text content in the booklet, or some images, or to change layouts based on demographic data known about the recipient. AI systems learn from data, and can evolve as datasets expand, or based on feedback from its output over time.
“Automation is software that follows pre-programmed rules. Artificial intelligence is designed to simulate human thinking.”
— Venture Beat
Five Flavors of AI with InDesign
True AI can be used with InDesign many ways. Here are five high-level possibilities:
Existing AI features of InDesign
To begin with, there are already AI features in InDesign, and there are more coming, both through the core product, and integrated Adobe technologies such as Adobe Firefly (currently in Beta). As of Today, AI features built into the product include:
- “Content-aware fit,” which lets you place graphics into frames based on the ability of AI to “see” the meaningful content in an image.
- “Find similar fonts,” which uses AI to find fonts similar to a selected font.
- “Subject-aware text wrap” which can identify clipping paths in an image and wrap text around them.
- Auto Style (available as a technology preview).
One possible use of AI in InDesign workflow is to render AI-generated text. Tools like ChatGPT and Jasper can generate text content, which is then easily placed into InDesign, in manual or automated fashion.
The downside of this is that we’re likely to see billions of publications that were mere echoes of human thought. In general, with AI, human guidance is still required to use it with integrity and efficiency. Rather than thinking of entire books or magazines coming out of AI, AI could be used to generate a call-out, or to suggest content or content variations. As we see with Adobe Firefly, in a serious workflow, AI handles a component of the process, or aids in a specific step in the workflow. Generating an entire piece of content from pure AI with no human review can easily lead to dumbed-down content that is a disservice to the world.
When generating text, or really AI content of any type, two sides of the process will have a huge impact on the results:
- Prompts: you can get very general content, sometimes totally incorrect, if you ask too general a question. It is good to narrow your query down as much as possible, and attempt various prompts, comparing results.
- The knowledge base of the AI application. The answers you get back are entirely based on exactly what the AI engine has learned.
Graphics can be auto-generated from prompts, as well as the combination of uploaded images and prompts. These can be placed in Adobe InDesign as graphics, manually or through automation. Both DALL-E and Adobe Firefly have robust image generation features. Adobe may have the upper hand here, as they can work inside of graphic files, with an awareness of layers, blend modes, etc., which can let human designers intervene and add finishing touches. Expect AI to be the primary focus of Adobe in evolving their products, across the Creative, Document, and Marketing clouds..
Document templates can be generated in various ways. This is actually a subtle thing: perhaps Adobe will make a template-generator specific to InDesign, but in our work we have found that the InDesign concept of “template” is very abstract and low-level. To effectively manage a template-driven publishing workflow, higher-level template concepts on top of the InDesign application lend themselves to automation, whether human or AI-generated.
AI Content Adaptation
If you look at the Adobe Marketing Cloud, you will find a fast-evolving set of AI tools concerned with targeting content, based on knowing who is going to read your document. These can include translation/localization, resizing/reformatting for specific output formats, tailoring content to reading-level or accessibility requirement, or data about the buying habits of the target recipient. Every dimension of the content could be tailored to the individual recipient or an audience segment.
Do you know of projects using AI and InDesign? Or ideas for what might be done with the two? Please mention in the comments and we can add them to this post.