Fahrenheit AI is an Ink Kitchen production in collaboration with Notorious R&D, to explore AI and its impact on the apparel embellishment industry. In this installment, we take a look at how AI models like Midjourney and Idiogram are handling text generation, examine the growing issue of deepfakes being used in politics, and walk through the process of creating an AI agent to automate blog writing.

When it comes to generating text along with images, Idiogram currently outperforms Midjourney. Using the phrase “Hammerhead Print” from an apparel design, Idiogram was able to produce images with the text spelled correctly and easily isolated. Minor adjustments allowed introducing stylistic effects like graffiti textures while maintaining legibility. Some generations had misspellings, but these could potentially be fixed in post-processing.

Midjourney struggled more with spelling the word “Hammerhead” accurately when generating text. While it produced some creatively stylized results, only one example had the word properly spelled. Midjourney also seemed to heavily tie the text to the imagery, limiting flexibility. For now, Idiogram appears to be the better option for apparel decoration artists looking to generate clean, editable text effects.

In the realm of politics, deepfake technology is becoming increasingly concerning. What was once mainly used by bad actors to falsely depict politicians is now being employed by politicians themselves. Last year, the Mayor of New York used deepfakes to generate robo-calls in multiple languages to promote city venues. More nefariously, candidates in elections have begun using AI-generated imagery to make opponents look weak or bolster their own images.

As deepfake capabilities grow more advanced and accessible, we may see them employed in apparel decoration for things like customizing mascots or branded characters. However, ethical considerations around consent and representation must be weighed carefully.

To help automate our own content creation workflow, we built an AI agent using Mind Studio to generate blog posts based on our Fahrenheit AI video transcripts. The process involved:

1. Defining the agent’s prompt, goals, and instructions
2. Providing it data sources (transcripts) and examples of our desired writing style
3. Choosing a language model (Claude performed well for our needs)
4. Setting up automations to query the data, apply the tonal style, and output a blog draft

The resulting agent was able to create coherent 700+ word posts summarizing the core AI topics we’ve covered, like pricing tools and manufacturing quality control. With some additional refinement of inputs and a combination of outputs, we now have a functioning blog writing assistant.

While AI text generation and agents can be powerful tools for apparel decorators, it’s important to keep ethics, consent, and branding integrity in mind. We must continually evaluate the implications of these technologies.

Check out the Fahrenheit AI playlist on the Ink Kitchen YouTube channel to learn more about leveraging AI in your apparel embellishment business.

New episodes exploring the latest tools and techniques are released regularly.Check out Episode #13 to explore these topics in depth.