I found a website that can use the image generation AI ``Stable Diffusion 2.0'' for free, so I compared the generated result with ``Stable Diffusion 1.4'' Review
Version 2.0 of AI `` Stable Diffusion '' that generates images simply by entering text (prompt) was officially released on November 24, 2022. A website where you can easily try Stable Diffusion 2.0 was released, so I actually generated an image and compared it with the conventional Stable Diffusion 1.4 generation result.
Stable Diffusion 2 |
When you access the link above, you will see a screen like the one below.
To generate an image, enter the prompt in the input area at the bottom and click 'Generate' on the right. This time, I entered the prompt 'A lion wearing a cowboy hat' as shown in the input example above.
The generated result looks like this. I was able to output an image of 'a lion wearing a cowboy hat' in about 30 seconds. Click Generate again if you want to generate again with the same prompt.
The generated result looks like this. This time, we were able to generate a monochrome lion image.
Below is the result of generating 6 images with Stable Diffusion 1.4 using the same prompt. Images generated with Stable Diffusion 1.4 have problems such as ``not wearing a cowboy hat'' and ``rough expression of fur'', and you can clearly see the quality difference from Stable Diffusion 2.0.
Next, 'girl with long pink hair, instagram photo, kodak, portra, by wlop, ilya kuvshinov, krenz, cushart, pixiv, zbrush sculpt, octane render, houdini, vfx, cinematic atmosphere, 8 k, 4 k 6 0 fps Below is an image generated with Stable Diffusion 2.0 by entering the prompt ', unreal engine 5, ultra detailed, ultra realistic'. In Stable Diffusion 1.4, the problem of ``human head is cut off'' occurs frequently, but in Stable Diffusion 2.0, the subject fits perfectly within the frame.
Below is an image generated with Stable Diffusion 1.4 by entering the same prompt. You can see that the outline is blurred compared to the image generated with Stable Diffusion 2.0.
「1girl, solo, smile, bow, jacket, :d, controller, hairband, holding, bowtie, bangs, blazer, shirt, purple eyes, open mouth, school uniform, looking at viewer, game controller, purple hair, upper body, Blue jacket, holding controller, long sleeves, short hair, holding game controller' and generating an image with Stable Diffusion 2.0 looks like this. The shape of the fingers and the shape of the controller are unnatural.
The generated result with Stable Diffusion 1.4 looks like this. Only the lower right one was able to keep the instruction 'girl with purple hair with a controller' while keeping the face within the frame.
Next, the result generated with Stable Diffusion 2.0 by adding the phrase 'anime style' to the above prompt to generate an illustration-like image is as follows. The image looks like a life-size board with lighting, but I followed most of the prompts and the outlines are clear.
The generated result with Stable Diffusion 1.4 is as follows. After all, Stable Diffusion 1.4 is not good at fitting the face into the frame, and the outline tends to be blurred.
To generate high-quality images that meet the prompts in Stable Diffusion 1.4, “Generate images with multiple seed values, choose a seed value that can generate high-quality images, and generate hundreds of images with that seed value. It is necessary to perform the work of 'selecting from', but with Stable Diffusion 2.0, even with a random seed value, it was possible to quickly generate a high-quality image as instructed by the prompt.
in Review, Software, Web Application, Posted by log1o_hf