No topic is currently getting as much attention as image and text AI - most notably ChatGPT. Curiosity coupled with our great love for data and technology is what drives us - which is why we too have turned our attention to AI tools like ChatGPT, Neuroflash, Midjourney, Dall-E etc. from day one. As a result, we've been able to identify several use cases for integrating image and text AI in everyday agency work.
Anyone who wants to work with AI should be aware of one thing: ChatGPT is neither a psychic oracle nor the (anti)hero of a sci-fi utopia (this is true for all AI models), but "only" a language model that uses sums of data. Unpopular Opinion, but: We do not feel threatened by artificial intelligence - especially in the field of creative work we are convinced to be and remain irreplaceable.
AI doesn't work against you, but - if you know how to use it properly - for you.
Nonetheless, and to clear this up once and for all, we checked directly with ChatGPT:
Applications and use cases for text AI
For AI to play a complementary role and work for us, we need one important skill: accurate prompting. In other words, the skills to feed the AI exactly the information that leads to the desired result. As always, the quality of the result depends largely on how precise the briefing is.
Five tips for creating prompts for ChatGPT
1. Give context: Topic, purpose of the request, for which target group should the answer be understandable?
2. What perspective should the AI take? E.G.: Answer from the perspective of a UX designer.
3. Provide guidelines on text type or length: E.G.: Compose a tweet.
4. Use predicates, e.g., regarding tone.
5. Provide examples or data.
So if you know what you want, these tips will give you good and helpful answers. But perhaps we'll take a step back and show where text AI can be used at all or usefully in a professional context.
How to use text AI in marketing in a meaningful way
1. Automated text creation: text AI can be used to automatically create texts (or get clues about their content or structure), e.g. for press releases, blog posts, social media posts or this list.
2. Sentiment analysis: Analyze sentiment in texts, e.g. customer reviews or social media posts, to measure customer satisfaction.
3. Text mining and data analysis: Text AI systems can be used to analyze large amounts of unstructured text data (e.g., emails, documents) and extract relevant information.
4. Keyword optimization: Automatically optimize content based on keywords by embedding them in the text.
5. Competitor site analysis: A text AI model can be used to analyze the content of competitor sites to identify optimization opportunities.
6. Writing meta descriptions and titles: Create meta descriptions and titles for pages that are optimized for search engines.
7. Generating content ideas: automatically generate content ideas based on user search intent.
8. Identify subtopics: To further optimize website content for search engines, use text AI to identify subtopics that current content doesn't yet cover or doesn't cover enough - but could be relevant to searchers.
9. Create FAQs: ChatGPT or other text AI models can be used to automatically create FAQs based on the content.
10. Translations: Text AI can translate content into other languages.
Really exciting content, use cases and tips in the application is provided by ChatGPT on LinkedIn.
The use of image AI in everyday agency life
Even with image AIs like Midjourney or Dall-E, the more specific my query, the better the result (in terms of meeting my expectations). Of course, you can also use the AI as inspiration tools and just see what it pours out with very little input.
Five tips for creating prompts for Midjourney
1. Structure of the prompt: It's best to describe the main composition before the first comma.
2. Specify enough attributes: After the first comma, specify attributes such as style, elements to include, colours to use, etc.
3. Use setting phrases: At the end of the prompt, for example, request the latest AI engine with --v 4, select the quality level with --q 2, or achieve rectangular results with an aspect ratio other than the default 1:1 with --ar 3:2 or --ar 2:3.
4. Include link to an image: Adopts its contents and can be customized or modified with additional attributes.
5. Trial-and-error: Don't settle for the first result, change sentence structure, attributes, style, etc. to get the perfect result.
As exciting as all this is, critical voices have been raised in the field of image AI in recent months. We asked ChatGPT to list the main pros and cons for us.
Pros and cons of using image AI
1. Facilitating the creation of art: image AI can help artists create complex visual effects faster and easier by automatically taking over certain steps of the creative process.
2. Expanding creative possibilities: By using image AI, artists can develop new techniques and methods that they could not use before, allowing them to shape their work in new ways.
3. Legal uncertainty: Since image AI technology is still relatively new, there is often uncertainty about how copyright law should be applied to generated images.
4. Use of copyrighted images: When image AI systems are used to generate images, they may use copyrighted images
5. Displacement of humans: There is concern that the automated creation of art by AI may limit the capabilities of humans. Therefore, there may be a displacement of human artists.
With these aspects in mind, the possibilities for the commercial use of AI-generated images are certainly limited. Nevertheless, we've found a variety of scenarios for ourselves in which using image AI takes a lot of work off our plate - and helps us visualize ideas convincingly even when time, resources or stock image portals reach their limits.
How to use image AI wisely to support creative work
1. Visualization of unusual and detailed ideas in the course of an idea pitch. An example from one of our concept presentations: close-ups of elderly people's lips puckered to a whistle, with cinematic lighting. A request where even the biggest stock portals reach their limits.
Our case about the donated campaign for the Haus der Barmherzigkeit "Wir pfeifen auf euer Mitleid" shows how close this presentation comes to the photos produced as a result - see here.
2. Image optimization and adjustment: Found a stock image, but the wrong lighting or too serious facial expression? With the right attributes, these elements can be revised by the AI.
3. Storyboards: Visualizing film concepts. We always present film ideas with a visual component. There is not always time and/or money to hire an* illustrator. The AI can do this frighteningly well: Simply describe scenes and add attributes like "storyboard pencil drawing".
4. Abstract visualizations and artwork: for our own blog, we like to use Midjourney and go for abstract representations of the content, like in this post. Or we create artworks together with the AI that symbolize brand identity and brand values, for example - like in the NFTs we give away - see here.
5. As an inspiration tool: Whether hero sections of landing pages or product stages for product presentations on social media - Dall-E 2 and Midjourney provide layout ideas for initial design scribbles.
AI in marketing, advertising and content creation? Our conclusion:
If you know how, you can make AI work smart for you. However, it is still a long way from the point where the currently existing models can replace the (creative) work of an agency. That's why we're doing it the way we've always done it: We are looking forward to the change with all its exciting developments and will take advantage of the new opportunities that AI offers.