FREE Image to Text Tool (Prompt Whispering)
Here is a tutorial on Replicate which is a free image to text tool.
We are used to putting text into Midjourney or DALL•E 2 and getting an image.
Well here is the complete opposite, you put in an image and you get the text.
Now you might be thinking why is this useful?
Well, if you’re looking to replicate a certain style or piece of content, this is a great starting point for getting excellent keywords that the machine is going to understand.
So let’s dive in and I’ll show you how this free image to text tool works.
Free Image to Text Tool: Upload an Image
So I’ll leave the link to the site in the description below
And what happens is you upload a photo you can drag and drop and put one in here.
The example given here is a cat in a suit.
And the machine then outputs a number of keywords that it thinks relate to the content inside.
So we have here a cat wearing a suit and tie with green eyes, a stock photo by Hanns Katz, pixels, furry art, a stock photo, creative commons attribution, and quantum wave tracing.
So we’ve got some interesting words inside of that as well as quantum wave tracing, which I believe is some sort of physics equation.
Create an Output
What we can do with this is we can put this right inside of Midjourney and create an output.
So we literally can copy all of these, go into Midjourney, type in: imagine and paste those in there.
And wait to see what we get.
So here is the grid of four that I received from Midjourney.
You can see it’s done an excellent job at imitating the content, though the quality is slightly more painterly or Illustrated.
I upscaled one of these and you can see we’ve got a pretty good one like this if you’re looking to get this sort of subject.
Use Photorealistic Algorithm
Now I went one step further and used the new photorealistic algorithm available in Midjourney.
And I’d have to say the results are improving remarkably.
So here you can see the original photo and the one that I created in Midjourney.
And so, this is an extremely useful tool if you’re looking to get a very specific look.
So I’ve also tried inputting an image I created in Midjourney, for which the prompt was classical space opera.
And it outputted these keywords a group of people standing in front of a piano painting by Wadim Kashin.
So you can see it has identified some of the key elements, but it has missed out perhaps the cosmic context here.
Take a Photo
So another cool thing that you can do is you can take a photo with your webcam.
I went ahead and took a photo of my face here.
And you can see that it has output a woman holding a piece of pizza in her hand, a stock photo by Bedwyr Williams, Behance, net art, demonic photograph.
Well, I say, androgynous!
So you can see it’s absolutely got a ridiculous output here.
Let’s see what Bedwyr Williams is like.
So we can see some of the works by Bedwyr Williams.
I think maybe it’s picked up a little bit on a pose like this.
So of course I went into Midjourney and decided to create this woman holding a piece of pizza.
This androgynous woman.
And this is what it has outputted.
I’ve upscaled one of these.
And here I am, this is what a computer thinks I am.
I am a woman holding a piece of pizza.
I believe the piece of pizza relates to the slight view of my pop shield on the microphone that I have here.
So after feeling a little bit offended by the machine, I decided to put a more natural photograph of my face in.
And once again I am a woman holding a plate of food in her hands, but this time it’s a screenshot by Esther Blaikie MacKinnon.
So if we have a little look at Esther, and the work that she does.
She was a Scottish artist known for her paintings.
And I think possibly it’s picked up on some of the tones.
The more muted neutral tones that I have in this photo because the style for me obviously is a photograph and these are all paintings.
It does not necessarily relate.
Put Keywords into Midjourney
So once again I put these keywords into Midjourney.
This time with the photo-realistic filter and this is what it comes out with.
I can actually definitely see the likeness here.
This does look a little bit more like me.
And a kind of interesting photo, so, thank you machine.
Try Different Types of Images
So I went on and I wanted to try a few different types of images to see how they performed.
This was the image I used here.
An output it gave was a group of bananas sitting on top of a blue surface, a stock photo by Helen Thomas Dranga.
And Helen Thomas Dranga is a British American painter who made paintings of Hawaii.
So perhaps it’s picked up on the tones and the context.
But once again this is a painter rather than a photographer that it has identified.
And yet, I went on and inputted this into Midjourney.
And these were the two best results that I got out of it.
The first one was with the photographic filter, and the second one was with the standard algorithm.
With this standard algorithm, I also included the original image as a prompt.
And this is what it came out with these were the four options it initially gave.
I then tweaked the prompt a little bit and added equally spaced, because it seemed to put the bananas in a higgledy-piggledy arrangement.
I also changed a group of two bananas, because it initially started keeping out these bunches of bananas.
And I have to say that the results are mixed here.
It doesn’t quite have the same feel as the initial image where the strength of the composition and the work here is really based on the arrangement of the bananas being random, yet consistent.
All the bananas here are not quite doing that.
They typically just sort of been dropped and spread out.
So beyond this, there is a landscape attempt and the output of this was a painting of a mountain landscape with a river running through it.
A matte painting by Asher Brown Durand.
This is pretty similar, I would have to say that these are very like the original and have done pretty well.
So I put these back into Midjourney and these are the outputs I got.
You can see that it has done pretty well in the context, of creating this mountain with a river running through it.
And yet the style for me is slightly darker.
I think there is a lightness to the original image that it’s missing out on.
There is no brightness added inside of the keywords generated by the algorithm, something it has ignored.
Painting of A Woman’s Head
And here we have a painting of a woman’s head with multiple circles around her, a hologram by Alex Gray.
And it performed not so well on this one either, especially if you’re looking at the type of image that’s been created.
There’s a very psychedelic feel to this and the output that it gave and none of them were really matching the content.
So overall, this free image to text tool is useful for a number of reasons.
Firstly, it helps you identify keywords, which you can use to create certain content.
It can also help you find artists’ styles, which may relate to the work that you’re trying to create.
And it’s a lot of fun to see what the computer thinks is inside of an image, especially if you put your own face inside.
I can also recommend this if you’re trying to identify who an artist might be from a piece of work.
So it’s almost like a Shazam for artwork.
I hope you enjoyed this video.
Let me know how you find this in the comments.
And if you’re looking to learn more about Midjourney, do check out my course in the links below.
I hope you have a delightful day.
If you enjoyed this video on free image to text tool, you will definitely like the following articles too.