Photo Hunt
Finding just the right photo for a presentation
For years, I’ve been building presentations for teams, clients, conferences and beyond. One thing we’ve all experienced when sitting through those slide shows is that the right image is way more powerful than huge blocks of 12 point font. For the past decade plus, it’s been a chore to comb through our own photos, Google Images (those that give you rights to use), or visit stock image sites like Unsplash (some good free imagery) or Getty Images (expensive professional shots).
With the rise of AI in the image generation space, it was only a matter of time when image creation tools were good enough to add to your tool-kit to help create dynamic presentations. I decided to try out several when making my last presentation that was geared around annual goals for the team. Below I will walk through the pros and cons with examples for each.
Goal: Show a team climbing a mountain at the end of a long hike
Prompt Used: climbing team on a mountain with the sunset in the background
DALL-E
Access: OpenAI Image Generator Service
Cost: Starting at $20/month
Process: Visit website, sign up and pay and then insert the prompt and it provides you with several image outputs.
Results: For my purposes, the people and imagery were a bit too cartoony and not realistic enough. I probably would not use them in most presentations but could be used if it went with the theme overall.
Examples:
Midjourney (v6)
Access: via Midjourney chatbot on discord
Cost: Starting at $8/month
Process: Sign up/into Discord, add Midjourney server, pay for a subscription and then enter a prompt.
Results: When changing the model to v6 the results were fast and much more realistic than the DALL-E examples above. I would feel comfortable using these in many presentations.
Examples:
Fooocus (based on Gradio + Stable Diffusion)
Access: Downloaded from Github
Cost: Free
Process: Download Fooocus from Github and install. This is a bit tricker for Mac users and I had to go through some other hoops outlined in the GitHub article in order to install. Once installed you need to run some basic Python script to update and start the engine on your computer. This can seem intimidating, but it was actually a fun process to go through and a learning experience. Once that is done, a GUI that runs in your browser on your machine opens. You can enter a basic prompt and then the images will generate. Because it is using your local machine to do so, it does take quite some time (30-60 minutes in my experience).
Results: The images themselves are stunning. I would have a hard time telling these weren’t real photos and I would definitely (and did) use these in my presentation.
Examples:
Unsplash
Access: unsplash.com
Cost: Free for some or $7/month for Unsplash+
Process: Go to their website and search for images of people climbing a mountain
Results: For this purpose I did find a few good pictures that were available, but wasn’t able to tweak what they showed, change what the aspect ratio was, or alter other attributes as I would be able to do with any of the above tools.
Examples:
Your Own Pictures
Access: Wherever you take pictures and back them up
Cost: Priceless
Process: Get to the right location, take amazing pictures, back them up to the cloud service of your choice.
Results: As good as you can make them!
Examples:
Conclusion
I’d love to be able to use my own photos whenever possible as it makes it more personal and you gain the experience of the moment as well. Unsplash also gives you some solid options if you want to stick with non AI generated photography. However, it’s not always possible to match what you need to tell your audience. For that, I believe AI is going to produce the most custom tailored results. Here are the winners across several key categories:
Cost: Fooocus - Free to use as it uses your local machine for compute power, but the time to set-up and time to create images could be a barrier to entry
Quality: Fooocus - I definitely believe that the stable diffusion data set and prompt+UI tools they give allow you to create the highest quality images at the moment.
Speed: Midjourney/DALL-E - Once you are registered and set-up both of these tools produce images in seconds not minutes
Ease of Set-Up: DALL-E - It’s a website that all you have to do is register and pay and type a prompt. No messing with downloading programs/running python scripts (Fooocus) or setting up a discord server (Midjourney)
Ease of Use: Midjourney / DALL-E - Once you are set-up both of these are easy to use to create images based on prompts and alter them as needed
Of these AI image generators, I would use the Fooocus and Midjourney images in my presentation. The DALL-E images I created weren’t quite there yet. If I had enough time, or didn’t have the budget, I’d go with Fooocus, but the time to process and set-up are barriers. If you need something more quickly and are willing to spend the money, I’d lean on the latest version of Midjourney for your task.