键盘摄影速成：多款AI人工智能绘画工具试用及推荐

A variety of AI artificial intelligence painting tools trial and recommendation

Addicted to AI painting, I started the stage of keyboard photography, and I only needed to type a few description sentences every day to create a few fake photos.

On the one hand, AI painting can make your boundless imagination come true, and it can be easily transformed into pictures. On the other hand, it can constantly stimulate your imagination, so that scenes that you never dared to imagine before can be generated at will. , As if after having a magic weapon, the discovery ability has increased.

From the very beginning of the DD platform, you can basically generate more than a dozen ai paintings every day, each of which takes 15-20 minutes, and then the free GPU usage is used up and you will rest. The efficiency is very low, and then start looking for The engine with better effect and faster speed has generated more than 800 pictures from the first dozen pictures to now at most. From only one or two scene pictures that can be used in a few days, now you can freely synthesize each picture. A close-up portrait of the real. Intelligent tools can greatly improve the efficiency of secondary creation.

At present, there are many AI painting tools on the market. Here I will share the tools I have used recently for your convenience.

DALL·E

AI painting tool developed by OpenAI. Create highly realistic images in minutes with this tool.
After the new invitation is registered, there is 50 points of computing power, and 50 sets of AI pictures can be generated.
Advantages: fast speed, four pictures at the same time, small distortion. Very convenient for secondary creation. You can upload reference images for the second creation of the same series.
Disadvantages: It does not support China, and the invitation to register and verify the mobile phone number needs to use a foreign mobile phone number (there is a platform that can be activated on behalf of it). This is the one I want to renew the most, but it does not support domestic credit cards.

Internal testing website: labs.openai.com/waitlist

Tool URL: https://openai.com/dall-e-2/

Midjourney

Midjourney is carried on the Discord channel. You can have a Discord account in advance. This is one of the most practical AI painting tools. The Ai painting that won the award not long ago was made with this tool
Open registration. After registration, you will enter the Discord channel. Newly registered users have about 25 experience times. After running out, they need to pay, about ten dollars per month.
Advantages: convenient, fast, generate four styles at the same time, and then continue to generate according to the style, and finally generate satisfactory pictures, paying users can also generate high-definition pictures.
Disadvantages: When the graph is generated in the channel, it is a bit messy and will be covered by other messages.

Tool URL: https://www.midjourney.com/

Beta URL: https://discord.com/invite/midjourney

Stable Diffusion

The AI painting artifact, like Midjourney, is also used on Discord. But his own platform can be generated directly, so it doesn’t feel as messy as Midjourney. From the training results, his drawing effect is slightly weaker than the previous two, the main reason may be the recognition and training of sentences. question.

Newly registered users will be presented with a test fee of about two dollars. Each time a picture is generated, it may cost about 0.1 yuan depending on the size of the picture. The more complicated it is, the more it will be consumed. Support image upload for secondary creation, which is very convenient for secondary creation.

Advantages: convenient and powerful, the effect is shocking, and it can be registered multiple times.
Disadvantages: The picture needs to be saved manually, the large picture only supports 1024

Official website: https://stability.ai/
Test address: https://beta.dreamstudio.ai/home

Disco Diffusion

Previous article “Poison Mirror: Am I an Artist?” But this is really not done by humans. . . “Introduced, it’s a little more complicated, and it has good support for scene generation. For some high-definition portraits and animal close-up pictures, it’s worse, and it takes a lot of time to get good results. There is also a disadvantage of being slow

Homemade 6pen.art

The domestic ai painting tool can be tested for free, but the queuing time has reached three hours. Since only two tests were tested, the effect was not good. The server may be a Google cloud server. The biggest highlight is that the test image size has reached Close to 10m, the only AI that generates a big picture.

URL: https://6pen.art/

Baidu’s Wenxin Yijie

Baidu’s AI painting platform is free, fast, and accessible in China. It also supports multiple model generation and multiple size options. However, it cannot be used as a second-generation tool at present, but as a toy.

The style is the way of choice, which limits the style of the work on the one hand, but also the idea. There is an advanced custom version setting, but the usage limit is relatively strict. When there are many times, three copies are generated, and when there is less, one copy will exceed today’s limit.
From the perspective of image generation, details and sentence recognition are not satisfactory. If the generated images need to be downloaded, they need to be reviewed by a machine, and many inexplicable cannot be downloaded.
The test account can be about 30 per day, but it seems that the browser can be opened more. Custom supports 2048 resolution patterns, does not support uploading pictures for secondary creation, generally speaking, you can play with nothing.

There is another one, among so many ai painting platforms, only Baidu has added a huge watermark in the lower right corner, which is really big.

https://yige.baidu.com/