How to chat about Images with AI (MiniGPT-4)
If you are all about Generative Ai tools at the moment you’ll be pretty curious about a brand-new Ai that allows you to upload images and ask questions about them. Simply upload the image that you want to chat or ask about, ask some questions and you’ll get a ton of information. It’s still in beta but is fully open-sourced and looks really promising.
Related: How to use Open Assistant - An Open Source chatGPT alternative.
Over the last 6 months, we’ve seen Ai go from a crazy sci-fi idea to actual working tools anyone can use. We’ve got Midjourney, Adobe Firefly, Sable Diffusion, DALLE, BlueWillow, etc for creating images. While ChatGPT, OpenGPT, and Google Bard are creating text content. However, there’s only one tool that is combining both of these aspects, MiniGPT-4. MiniGPT-4 is an AI tool that allows users to upload an image and receive information about it through natural language generation.
Putting it simply, MiniGPT-4 is a generative Ai system that combines two different types of advanced computer models: one that can recognize and understand images, and another that can generate language in a natural and coherent way. This process involves training the MiniGPT-4 to align the visual features of the image with the language model. This means it can generate a description or story or answer questions based on what it sees in the image.
How do you chat and ask about Images using MiniGPT-4? Chat with generative AI using MiniGPT-4.
- To use MiniGPT-4 head on over to the website
- Once you are on the main page upload an image that you would like to talk about or ask about. At the moment the service is super slow as you will more than likely have to wait in a queue but this is standard for small open-sourced start-ups
- Once your image has been uploaded. You’ll be able to ask a question about the image. Again this process takes some time and while it shows a rough estimate of the time, that is almost always wrong.
TIP: If you want to do multiple images I suggest using multiple tabs, this will allow you to get more things going and make the wait times less annoying.
- You can continue asking questions about the image for as long as you want, you aren’t limited to a single question for each image.
- For example, you can even upload something such as an image of a tree with sick leaves and MiniGPT-4 will attempt to diagnose the issue and give you some reasons or solutions. Just use your brain when doing this and don’t automatically take everything it says as the absolute correct answer.
As we’ve mentioned a few times there are some issues with this tool at the moment but the concept is awesome and although MiniGPT-4 is the first of its kind I’d expect to see this exact same system added to tools like ChatGPT. OpenAi already has a generative Ai for images (DALLE) so merging DALLE with ChatGPT wouldn’t require all that much effort on their part. Anyway, that wraps up this guide, if you’re into all kinds of Ai stuff make sure you check out our YouTube Channel as well, we’ve got a fair bit of interesting Ai content over there.