Although 2022 is not the time when the technology of AI drawing was born, it is the first year of drawing that has exploded so far.
AI drawing (AI painting) is based on text2image, which is a kind of cross-modal generation: it refers to the conversion of one mode (text, image, voice) into another mode. modalities while maintaining semantic consistency between modalities.
With such a hot AI drawing, where can I generate stunning pictures? Today, I will recommend 7 AI painting tools to everyone, and introduce the functions and prices of each tool in detail. I hope you can find the AI painting tools that suit you after reading this article.
1、PicUP.AI
After repeated testing, PicUP.AI is finally officially launched! Compared with other AI painting artifacts, PicUP is faster and has better generation effects. The most important thing is that the whole site is in Chinese, and a picture can be generated every 5 seconds.
Users can use their imagination and input a text description to generate a 1:1 picture. There are also 15 painting styles to choose from. In addition to basic anime and realistic styles, there are mainly some special styles, such as surrealism, gloomy style, cyber style, ethereal, sci-fi, steampunk, etc.
In addition, PicUP.AI also has the function of generating pictures from pictures, which is a further attempt based on Vincent's pictures. Just upload a reference picture to generate a picture with a style similar to the picture, or create a picture based on the original picture. Add your own ideas on it. If you are not satisfied with the generated results, you can also regenerate.
Every image you generate can be published online and viewed in the "Gallery", allowing more users to see your work. Currently, free downloads with watermarks are supported, but 5 image processing points are required to download images without watermarks.
2、FROM-E 2
Both DALL-E and the upgraded DALL-E 2 come from the OpenAI team. This top international AI research institution co-founded by Musk announced the Image GPT model in July 2020. It will be a Transformer that will make a breakthrough in natural language processing. The model introduces image completion and generation tasks. Not only did it open source a new deep learning model CLIP (Contrastive Language-Image Pre-Training) in January 2021, a state-of-the-art artificial intelligence for image classification, but also launched a new AI model DALL·E. Just input a text description, and DALL·E can draw a series of alternative images that meet the requirements. It can be said that this is the first platform to realize "drawing from text".
DALL-E 2 is its upgraded version. It is characterized by realistic style, simple enough operation, high degree of completion, and fast enough to be used as a search engine: 10 pictures (1024 × 1024) can be generated within 60 seconds, which can be extended and changed infinitely, and can even be erased and regenerated partially.
New users can generate 200 pictures per month for free, then 60 pictures per month, and 460 pictures for $15 later.
In terms of copyright, OpenAI, the organization behind DALL·E 2, has listed several strict restrictions: the copyright of image generation ultimately belongs to OpenAI; it is only for personal learning and exploration, not for commercial use, and cannot be used to make NFT; it cannot be published on social media too much Realistic face generation results will have the risk of portrait infringement.
3、Midjourney
Midjourney is the platform where the previously generated painting "Space Opera House" won the award against human painters.
It is characterized by a simple interface and a variety of options. Midjourney rides on the communication software Discord, enter "/image" in the dialog box, enter the description in English, and press the Enter key. This process is like chatting with AI. After 60 seconds, you can receive 4 rendered images in the dialog. If you are not satisfied with "Figure 1", you can click the "U1" button to add details, and press the "V1" button to extend the changes until you are satisfied.
Midjourney has a creative community, zero-threshold interaction and very good output results. From the output style, it is very obvious that some optimizations have been made for portraits and the style tendency is also obvious.
Each new user has the opportunity to generate 25 pictures. If you want to generate more pictures, you need to pay. The basic membership is $10/month and has the opportunity to generate 200 pictures; the standard member is $30/month and can generate unlimited pictures. .
In terms of copyright, if you are a free user, the copyright of the image belongs to AI, and after paying $30 per month, you can use the image for commercial use. But at the same time, if you make a profit of more than 20,000 US dollars, you need to give Midjourney 20% of the share. Because it is a paid business, MidJourney's iteration is very fast.
4、Disco Diffusion
If DALL-E 2 is good at realism, then Disco Diffusion is better at depicting atmosphere, scenery and various concept art.
Disco Diffusion can realize the most complex keyword description, and supports many self-setting parameters. It is a huge treasure trove to be explored. It can run, and it is free at this stage, but the operation is more troublesome. The entire webpage is full of codes, and the operation is difficult. Generally speaking, you need to wait for half an hour. If you stare at the screen, you will see that the image gradually becomes clear and detailed from being full of noise. During use, Disco Diffusion may prompt you to free up enough running memory on the computer, but because it runs on computing resources such as GPU provided by Google for free, it does not require high computer hardware for users, just open the browser and run it. In addition to only entering text to let AI play freely, you can also insert an initial image (Initial Image) in advance to constrain AI's creation.
The graph generated by Disco Diffusion can theoretically be used commercially. Its program is based on the MIT open source protocol. All Internet users can use, copy, modify and even sell the generated graph for free. But there are still risks. The risk mainly comes from the controversy that your descriptors will attract plagiarism.
5、Stable Diffusion
Stable Diffusion is considered to be the most powerful AI painting tool at present. It is completely open source, and there are many "variants" on the market, such as Waifu Diffusion, which is dedicated to generating two-dimensional portraits. It can quickly (calculate in seconds) generate a 512x512 image full of details, and only need a consumer-grade 8GB 2060 graphics card to achieve DALL-E 2-level image generation, and the generation efficiency can be increased by 30 times. The style is obviously more artistic, and it is not difficult to get started.
Charging standard: There is a free generation quota of 200 sheets, and then you need to pay to purchase points (the more complicated the generation, the larger the size, the more points will be consumed).
Copyright requirements: You can commercialize your own images, but if the images are generated by DreamStudio, they will automatically become CC0 1.0 authorization, so that the service provider Stability.ai can also process your images without paying or even passing through you Agree, and it will also become a general public domain royalty-free picture resource. If you deploy the open source Stable Diffusion by yourself and consume your own GPU resources, then the copyright belongs to you.
6、Google
As one of the first companies to study AI painting, Google launched Deep Dream as early as 2015. Recently they released two models in one go: Imagen and Parti.
Imagen's image generation has a diffusion model similar to DALL-E 2's, but feeds on a large AI language model — resulting in better image generation results from textual descriptions due to higher language understanding.
The new AI model Parti (Pathways Autoregressive Text-to-Image) attempts to use an alternative architecture (autoregressive) that more closely approximates the capabilities of large language models that predict appropriate new word. Parti applied this principle to imagery, with success. Parti can accurately translate long and complex text inputs into images, suggesting a better understanding of the relationship between language and topics.
Accompanying Parti's release is a blog post describing the process of creating images using Google's text-to-image model, which can be accessed here: https://blog.google/technology/research/how-ai-creates-photorealistic -images-from-text/
However, Imagen and Parti have not yet released a test version, and their specific usability remains to be verified.
7. Novel AI and others
If you want to try the two-dimensional style, then you must try NovelAI. The AI painting function charges: 10 USD monthly fee 1000 tokens, 25 USD monthly fee 10000 tokens, each painting consumes certain tokens. NovelAI is cc0 copyright, that is, public copyright.
In addition to these applications, more models and commercial applications are emerging continuously: NUWA-Infinity launched by Microsoft, Make-A-Scene launched by Meta and other platforms such as NightCafe Creator and WOMBO Dream.
In general, the materials generated by AI painting can be used for drafting or divergent thinking, and can also be a reference for materials, inspiration, abstract concepts, and color effects. Most importantly, AI painting tools with faster generation speed can quickly help painters and designers build multiple possibilities, providing more ideas and unexpected ideas.
What do you think of AI painting? Will you use these AI painting tools? Welcome to leave a message in the comment area.