One of our most requested features by far is having our ChatGPT website chatbots display images or screenshots. Whether it's to show visually how to fix something for support related questions, show pictures of the team behind a company, show pictures of the food you're getting a recipe for or show pictures of products, this is a popular request. Almost every single client asks for this.
I completely understand why. Pictures can help the user understand better, it can give them a visual aid or it can help you sell a product, for example if you're using our ChatGPT website chatbot for e-commerce. Simply put, for almost every single use case, having your chatbot display images can increase customer satisfaction, user friendliness, sales and leads generated. While we've had the ability to show images and screenshots for a while, it's been a manual and time consuming job for most chatbots. This means that most clients wanting this feature have needed a lot of help to set this up, and it's taken a bit of time to do so.
The sneak peak
We're getting ready to launch a new update that contains a lot of new features, new logic and a significant increase in quality. One of many things that is getting an update is the logic on how snippets (pieces of training data) are formed, as well as how a question is matched with relevant snippets. Part of this is due to OpenAI's newest update, that you can read more about here. I immediately recognized this as amazing news, but I had no idea just how amazing until we started working on implementing the new logic to make use of this update.
What is actually changing?
So, how does this affect displaying images? As mentioned, it's a feature we've had for a while, but it's been slightly hard to use and most clients have needed help to implement it. The creation of the chatbots is a creative process, but the images feature has been a strictly logical feature that has needed to be done in a specific way to work correctly. With our next launch, this is a thing of the past. Due to how our crawler scrapes data with the new launch, it's no longer going to be a manual job for most chatbots. Our crawler already has amazingly high quality, but we've actually been able to further increase the quality. For the exceptions (say for example if the links to the images are broken), adding images and screenshots manually after crawling is much much easier and faster. Simply add the link using the markdown formatting to the relevant snippet, which is much easier to find now!
The last sneak peak you'll get today is that we're finally going to be able to have the GPT models (the models that remember conversations) use the markdown feature correctly for displaying links as well! While we were able to make it work in some chatbots, it definitely wasn't perfect. With this new update, we're finally going to be able to deliver perfect markdown links even in the GPT models!
Of course, I'm working on new training material which will show all the details - and while I can't promise how long it'll take, I'm planning on writing more detailed blogs about it as well as making videos showing the process for everything.
Until the launch, feel free to test out our own chatbot to see how well it works - we haven't finished everything and it definitely needs more testing, but even so the quality has increased tenfold since we updated it. There is no specific date for the new launch just yet, as we need to do a lot of testing to make sure we don't launch it with any bugs, but we're working hard to launch it ASAP! Here are some example prompts you can test our chatbot with to see how it works:
- Show me images of the team members behind AINIRO
- Show me a picture of (Thomas, Tage or Aria)
- Show me a picture of the QR code to test the WhatsApp chatbot
- Show me a picture of Frank, your AI mascot
- Show me a picture of Hyper IDE
- Show me a screenshot of your ChatGPT chatbot in use
If you click on the picture, it'll be enlargened so that you can see details better.