google bard image

Google Bard’s Image Mode Can Create Websites From Sketches And Screenshots

A few months ago, ChatGPT had stunned the world with a demo of developing an entire website from a simple sketch on a napkin. While ChatGPT’s open version still doesn’t allow for this functionality, Google Bard can do all this and more — for free.

Google Bard can now accept images at inputs, which has led to people unlocking a slew of new functionality from the LLM. A Twitter user, for instance, was able to create a Swift app from a simple hand-drawn sketch of a website which had two buttons. Bard was able to understand the image, read the instructions, and then create a near-perfect Swift UI for the website.

Another user was able to create an even more complicated UI on Swift using Google Bard. A user drew a rough sketch of an app which calculated the tip amount. Google Bard was able to understand the instructions, parse the sketch, and create a fully-functioning UI.

Other people were able to enter a rough mock-up into Bard, and it created code for the whole website.

And Bard does a reasonable job of even converting an existing website’s screenshot into HTML and CSS.

But creating websites isn’t the only thing the updated version of Bard can do. Others have found more creative use-cases for the AI to be able to understand images. A user was able to upload an invoice, and receive the contents in a table format.

Google Bard can also convert HTML tables into text, essentially acting as a free OCR (Optical Character Recognition) provider.

Some people have used Google Bard to convert images of equations into Latex.

Others found more unique use-cases. Someone uploaded a picture of a stack of books in their house, and Google Bard was able to read the titles, find the ISBN numbers, and create a table.

And the use-cases didn’t end there. Someone shared a picture of a Physics free body diagram on Google Bard, and Bard was able to generate Python code to describe the system.

And incredibly, Google Bard was able to explain a joke from just a meme, correctly understanding the image, combining it with the text, and articulating why the image was funny.

Google might’ve been slow off the blocks in the LLM race, but it sure looks like it’s catching up with ChatGPT.