Jul 13, 2023 11:14:00

Adobe's image generation AI 'Firefly' can now be used in Japanese, so I tried it

While various image generation AIs such as 'Stable Diffusion' and 'Midjourney' are appearing one after another, ' Firefly ' developed by Adobe has attracted a lot of attention from companies as

an AI that can generate copyrighted clear images. I'm here. Since such Firefly began to accept instructions in Japanese, I tried to see what kind of image is actually generated by what kind of instruction.

Adobe Firefly prompt input now supports multiple languages including Japanese, German, Spanish, French, Brazilian and Romanian
https://blog.adobe.com/jp/publish/2023/07/12/cc-adobe-firefly-gluhwurmchen-luciernaga-luciole-vaga-lume-licurici

Image generation by Firefly can be tried as a beta version at the following link.

Adobe Firefly (Beta)
https://firefly.adobe.com/

After accessing Adobe Firefly (Beta), click 'Login' at the top right of the screen.

Log in with your Adobe, Google, Facebook, or Apple account. Since I want to log in with my Adobe account this time, I entered my email address and clicked 'Continue'.

Enter your password and click 'Continue'.

Then click 'Generate image from text'.

In 'Generate image from text', you can generate an image as instructed simply by entering the characteristics of the image you want to generate in the input field at the bottom of the screen and clicking 'Generate'. This time, I entered 'a bear playing a ball in the forest' and clicked 'Generate'.

After waiting for a while, four images were generated as shown below.

On the right side of the image generation screen, there is a 'content type' that allows you to specify the style of the image to be generated. Click 'Photo' to try it.

Then, a live-action image was generated.

I will try to create more images. First, generate a 'bear that makes salmon sushi'.

'Old man drinking coffee by the window, it's raining outside.'

When I added the information 'sad expression', it generated an image with a sad expression.

It's like this when it's a 'magical girl who summons a lot of donuts'.

It is also possible to specify the style of painting, such as 'a boy eating ice cream on the way home from school in a Japanese anime style'.

It is like this when it is 'painting of smartphone drawn by Van Gogh'.

To download the image, move the cursor over the image and click the download button OK.

As 'promoting AI transparency', you will be notified to embed authentication information indicating that it is an image generated by AI, so click 'Continue'.

The image is now downloaded. The downloaded image had a file size of 1.31 MB and a resolution of 1024 x 1024 pixels.

At the bottom left of the image is the Adobe logo and the words 'Not for commercial use'.

The image generated by Firefly cannot be used for commercial purposes at the time of writing the article, but the English help page for the image generation function 'Generation fill' by Firefly scheduled to be installed in Photoshop says, 'Commercial use of the generated image in the second half of 2023.' will be allowed.'