How to generate AI illustrations for free? Introduction to alternative paid services
To generate AI illustrations for free, you need to install and run the platform on your own PC. First of all, we would like to inform you that generating an AI illustration on a low-spec PC may take quite a long time, or may stop due to insufficient memory error, etc. In such a case, you should replace your PC with a mid-spec PC or higher. In such a case, please read this article as an alternative to buy a PC of middle spec or higher, or use a paid AI illustration generation service as described in the latter half of this article.
PC specs required to generate free AI illustrations
A PC of middle spec or higher is required to generate free AI illustrations. Please refer to the table below. For video cards, we recommend Nvidia 3000 series or later, but older cards can be used if they support the python package “pytorch”.
| Required Specs | Note | |
|---|---|---|
| OS | Windows 10/11+ 64bit | |
| CPU | Intel Core-i Series・AMD Ryzen Series | |
| RAM | 16GB or more | |
| GPU | VRAM 8GB or more 12 GB or more for LoRA Train | NVIDIA recommended Internal GPUs are not available |
Required specifications for SDXL and other high quality illustrations
This table shows the minimum specifications required.However, if you are using the SDXL model, which is capable of producing high quality illustrations, to produce large size illustrations, it is recommended that you have at least 16 GB of video memory for comfortable production. The process of generating AI illustrations involves entering “prompts” which are instructions and parameters for generation (generation → confirmation → correction → generation), so if the generation time is long, the process will not proceed smoothly and will be stressful.
For reference, the benchmarks for the PC used by DCAI at the time of writing are listed below.
- OS: Windows 11 Pro
- Platform: AUTOMATIC1111 Stable Diffusion web UI
- GPU: RTX3090(24GB) ASUS TUF GAMING
- CPU: Intel Core i9-12900KF
- RAM: 128GB
- Storage: M2. Samsung 980 PRO 1TB x2 RAID 0
The sample illustration (1344×768 pixels) was generated in 6.5 seconds at 3.50 it/s, and including AI upscaling (2016×1152 pixels) and face modification, it took 23.4 seconds at 1.31 it/s. (it/s | Iteration per seconds: number of steps calculated per second)
In addition, since the platforms and models required to generate AI illustrations download several gigabytes of data, a large, fast storage capacity and a fast Internet connection will make the process even more comfortable.
Introduction of a free generative AI illustration platform
To generate AI illustrations free of charge, you can download a platform that operates with a web browser and generate AI illustrations on your own PC by using “local generation. Here are some platforms that support local generation.
Automatic1111 Stable Diffusion web UI
The first one is AUTOMATIC1111’s stable-diffusion-webui, which is also introduced in DCAI. A1111 WebUI is a platform for local generation of AI illustrations and has over 130,000 stars on GitHub. The Gradio-based UI makes it easy to use even for beginners.
DCAI introduces everything from installation to usage, so if you are interested, please refer to the following article.
ComfyUI
“ComfyUI” is a platform for generating AI illustrations with a node-based GUI. It can be used to generate AI illustrations for a wide range of applications including images, videos, and music. Compared to A1111 WebUI, ComfyUI may be a little difficult to use, but it is a lightweight platform that can generate AI illustrations even on a mid-spec PC. The node-based GUI is also highly scalable and can be adapted to new technologies quickly.
ComfyUI is also introduced in DCAI, so if you are interested, please read on.
Stable Diffusion WebUI Forge
“Stable Diffusion WebUI Forge” is a platform that can be used with the Gradio-based UI. The platform is once again attracting attention with its 🔗support for Fulx.1. Originally a lightweight version with memory management and other features when A1111 WebUI was unstable, it is now a repository for developers to test new features of A1111 WebUI. If you would like to use the previous version, you can find the download file at 🔗Download Previous Versions.
SwarmUI
”SwarmUI” was originally named StableSwarmUI and developed by Stability-AI, the developer of SD1.5, SDXL, and SD3 models, but has now been taken over by mcmonkeyprojects. It is lighter than A1111 WebUI and can be used on a mid-spec PC.
Fooocus
”Fooocus” is a platform that generates illustrations simply by entering prompts, and although you can change some settings in the Advanced menu, you cannot add extensions like A1111 WebUI. So, for now, we recommend it to those who would like to try out the AI illustrations generated.
SD.Next
”Next” has a modern UI, and compared to A1111 WebUI, it can use ControlNet and other essential extensions in the initial installation state, so high-quality illustrations can be generated immediately after installation.
Paid server that can generate AI illustrations even on low-spec PCs
Although not the “free” service mentioned in the article title, this section introduces a paid service recommended for those who cannot enjoy local generation due to an inadequate PC. The service can be used on any environment where a basic Internet browser is available, so it can be used on low-spec PCs and smartphones. The difference between this service and the AI illustration generation service we will introduce later is that it uses an open source platform, which allows for a high degree of freedom and scalability.
Google Colaboratory [Google Colab]
”Google Colab” is a Python-specific cloud service provided by Google. If you install the platform we have just introduced, you can generate AI illustrations using a powerful GPU. It used to be free to use, but now we recommend using a paid plan because you will get a warning if you continue to use it for free. You need some knowledge to install it, but after installation, you can use it as if it were locally generated in your browser, so why not give it a try? We will introduce the installation method on DCAI when we have a chance.
RunComfy
”RunComfy” is a paid server that allows you to use the platform’s ComfyUI with the server’s powerful GPU. It is recommended for those who are stressed by local generation on low-spec PCs, as it allows free use of everything from custom nodes to models. Although there are some restrictions, you can also use it with a free if you are on the Hobby plan.
Paid AI illustration generation service that can be used on low-spec PCs
We recommend using a paid AI illustration generation service because it allows you to use the server’s high-spec GPU, which can be used in any environment where an Internet browser is available, from low-spec PCs to smartphones and tablets. However, please note that compared to open source local generation, it is not possible to freely expand the functionality of the service.
Midjournery
”Midjournery” is one of the most popular AI illustration services, and we recommend using Midjournery first if you want to generate high quality illustrations. There is also a service called ”Niji-journey”, which specializes in illustrations, and is recommended for those who do not need to generate photorealistic images or who mainly want to generate anime illustrations.
NovelAI
”NovelAI” is an AI illustration generation service specializing in illustration generation. It specializes in anime illustrations rather than photorealistic images, so those who are not interested in realistic AI are recommended to use this service.
Civitai Image Generator
”Civit AI” is introduced as a model sharing site, but it also offers AI generation services. One of its strengths is that you can use models uploaded to Civit AI that you have permission to generate. There are not so many services that allow you to use such a wide variety of models.
Tenser Art
”Tenser Art” is a service similar to Civit AI above. There is a free slot available, so feel free to use it. There is also an application, so you can use it from your smart phone.
Runway
From this point on, we will turn to paid AI illustration generation services that are strong in video generation. ”Runway” started as a Gen-1 service (video to video) and was upgraded to Gen-2 (image to video / text to video). The current version, Gen-3 Alpha, further refines the video quality with control models such as motion brush, advanced camera control, and director mode. It produces videos in a wide range of styles, from cinematic to animated.
KREA
”KREA” started with ‘real-time generation,’ which generates images based on rough drawings on a canvas, and ‘enhancer,’ which upscales uploaded images and videos. Later, the AI Video generation service was launched, and it has become an easy-to-use AI Video generation tool that can be edited using keyframes.
Luma AI
”Luma AI” generates high-quality AI videos by using the Dream machine. The generated videos are very realistic and were the hot topic of conversation when the Dream machine was first announced. Not many videos can be used, but there is a free slot available, so those who are interested can try it out.
KLING AI
“KLING AI” is the next generation of AI service from China, which provides AI-generated images using KOLORS models and AI-generated videos using KLING-developed models. KOLORS is a model that can be generated with Midjournery-v6 class quality. The models in the generated videos produce smooth videos with no sense of discomfort.
Haiper
“Haiper” is a service for AI-generated videos, funded by Google DeepMind. The service offers Text to video (generated from a prompt), Image to video (generated from an uploaded image), and Video to video (generated from an uploaded video). The quality of the video is medium among the services introduced here.
Kaiber
”Kaiber” is a service specializing in AI-generated videos. The audio reactive feature, which generates videos in time with uploaded music, is a feature that cannot be found anywhere else. However, the quality of the videos is of a medium quality, which is a bit old-fashioned for AI-generated videos.
Conclusion
In this article, We introduced how to generate AI illustrations for free. The second half was a paid introduction 😅. We hope you understand that there are two ways to generate AI illustrations, one is to use your own PC and the other is to use a server to do the calculation. If you use a paid server with a budget for a high spec PC, your budget will be enough for several years. However, I think that the high degree of freedom of local generation is also something that cannot be discarded. This time, we only introduced the system, but we hope to introduce how to use it at another time. Thank you for reading to the end.
