A site where anyone can try `` img2img '' mode that automatically generates pictures and illustrations as desired with just simple drawing and keywords with `` Stable Diffusion ''
Stable Diffusion ' that creates human-like images according to keywords has been released to the public, and a large amount of high-quality images are being generated. In Stable Diffusion, in addition to 'txt2img (text to image)' that generates images from character strings, there is img2img (Image to Image) that can generate new images not only from characters but also from images. Since the demo page ' Diffuse The Rest ' where you can try img2img of this Stable Diffusion for free has been released, I actually tried it.
An AI called '
Diffuse The Rest - a Hugging Face Space by huggingface
Below is the Diffuse The Rest page . In the center is a canvas for painting.
Click the black square on the far left to change the color, and click the round black to change the thickness. The following icons are buttons that allow you to switch to pen, switch to eraser, and erase all.
First, draw a suitable illustration on the canvas, enter the letters related to the picture at the bottom, and click 'diffuse the f rest'. This time, I drew a picture of tofu and entered 'tofu'. Waiting for about 2 minutes and 40 seconds for image creation ......
A picture like this was generated. At first glance, it's clearly not tofu, but the white body has a brown color that looks like grilled marks, and there are things like tableware, so if you say ``it's tofu,'' you'll probably be convinced. .
Then, for some reason Apple's corporate logo was generated.
Next, I tried to enter a sentence that combines words. Draw a picture like this: 'a man with a gun, the sun is rising, standing on the desert, skull lying on the ground.' Skull lying on the ground), airplane (airplane), a vulture is flying (flying vulture)' ...
A picture like a scene of a science fiction movie was generated like this. The sun, the desert, the man with the gun, and even the vulture are properly generated, but the plane and the skull seem to have disappeared somewhere.
Multiple images are generated, and you can enjoy several different patterns.
Now enter 'Astronauts in Space, The lush green earth is about to be attacked by extraterrestrials'.
Then, he adjusted the illustration like this.
Another generated image looks like this.
It is possible to generate a new image from a photo, so I will try this as well. First, upload an image from local storage to the canvas by clicking and dragging.
Pinot Fragrant Milk Tea '. If you enter 'chocolate ice cream' and 'pino' as a bad source ......
This time I used a picture of '
As expected, an image of Pino was not generated, but we succeeded in generating an image of chocolate ice cream with a similar color.
In addition, Reddit has also posted that it succeeded in generating the following image. This person said 'A man with golden armor, and mask, rises from the sands, a shiny golden magical staff in one hand, Artstation, Cinematic, Golden Hour, Sunlight, detailed, elegant, ornate, desert, rocky mountains, a big shiny white sun in the background, Illustration, by Weta Digital, sandstorm, Painting, Saturated, Sun rays'.