![[Tutorial]iOS, Android free AI real-life drawing, no need for a computer + make a real-life cosplay beauty photo](https://cdn.unwire.hk/wp-content/uploads/2023/03/AIDraw-01.png "[Tutorial]iOS, Android free AI real-life drawing, no need for a computer + make a real-life cosplay beauty photo")
Stable Diffusion, which is used in AI drawing, can be said to be a hot topic on the Internet recently. However, if you look at online teaching, they all teach you to set up a server on a computer, install a lot of files on the computer, and the computer needs a powerful graphics card. This time, the editor will introduce how to use the lazy bag to set up a private Google cloud server for free with one click to play Stable Diffusion. You only need to have a Google account, and then press two or three buttons to create your own private AI drawing production center. As long as you are on an iPhone, Android phone, tablet, computer, etc., you can play Stable Diffusion on your private server through a browser. Play AI graphics without asking for help, and cast spells freely without fear of being seen!
A brief introduction, this time I will introduce Stable Diffusion, which is used for AI drawing. Originally, it needs a powerful computer and a powerful graphics card to generate AI realistic beautiful girl images in a short period of time. Now we directly build a cloud virtual computer on the Google Colab service provided by Google, and directly use Google’s computing resources to install Stable Diffusion for us to create images of beautiful girls. Benefits are:
free
You only need to have a Google account to use the Google Colab service. Unless you are doing a huge calculation of an enterprise-level organization, it is still free to use.
It can be created with one click, no computer is needed, and mobile phones can be played
There are already “good-hearted people” on the Internet who have made a lazy bag, which is so convenient that you only need to press a few buttons according to the screen, and you can successfully create your own free Stable Diffusion server, that is, you can play AI graphics without asking for help! And since the free cloud computer provided by Google is used this time, you only need a mobile phone to operate the AI drawing interface, and you don’t need to worry about the settings of the computer yourself!
Free self-built Stable Diffusion AI graphics server in 3 steps
Choose your AI drawing model
Step 1
To play AI drawing with Stable Diffusion, you need the AI drawing model. And this website, you can freely choose the AI model.
URL: https://github.com/camenduru/stable-diffusion-webui-colab
▲Open the home page in the picture and scroll down to see
On this page, you can see different types of models, and each model can create different styles of character images, such as “anything_4_webui_colab” with a comic style. This time, the editor focuses on the model “chillout_mix_webui_colab” that can draw Asian female faces. Find it in the menu and click the “Open in Colab” button.
Start your AI drawing model on Google Colab virtual computer
Step 2
Click the link just now, and you will be sent to the Google Colab virtual computer screen shown in the picture (login to your Google account if instructed to do so). You can see that the instructions to install Stable Diffusion have been pasted automatically. Now all you need to do is press the execute button of the instruction set.
Step 3
The instruction set will be executed automatically. You just need to wait quietly for a minute or two. Please scroll this page to the bottom (smart phone users are recommended to scroll the page on the left hand side of the screen), and wait patiently until the words “Running on public URL:” appear on the screen. The URL shown here is the AI drawing maker created by the virtual computer for you, just click to enter. You’re done.
▲Click the link in blue next to the words Running on public URL:
▲You’re done, the Google Colab cloud AI mapping server is available
Beginners get started with AI drawing
If you’re an AI mapping expert, this is the end of the tutorial. But if you are a beginner of AI drawing and playing for the first time, let’s start with the introductory “spell”. Afterwards, this tutorial will also introduce the import of a cosplay setting file called “LORA”, which is used to dress up different characters and change faces for beautiful girls.
Step 1 Enter the spell
First of all, what is a “mantra”. The “Generate” button in the AI drawing interface is to create a picture of a beautiful girl, but if you just click it without any instructions, the produced picture will be as shown in the picture, which is not what you want.
The so-called “spell” is an instruction given to AI, directing it how to help you draw the picture in your mind. On the “Generate” button, you can see the “Prompt” command (that is, tell the AI what to do), and the “Negative Prompt” negative command (that is, tell the AI not to do anything) two fields, for you to enter to command the AI The “mantra” of drawing. Let’s try the mantra of the original heart! (Usage: Paste the following Prompt and Negative Prompt respectively to the corresponding two fields of the drawing interface)
Paste the following spell in the Prompt field:
(photographed using Sony A9 II (Mirrorless Camera), by photographer, photorealistic, 1girl, kpop idol, detached sleeves, long hair, best quality, (photorealistic:1.2), (hair ornament:1.35), jewelry, earrings, small breasts, torii, cherry blossoms, lantern light, depth of field, detailed face, face focus, (looking at viewer:1.25), shiny skin, no sleeves, smile, thick lips, thighhighs, east asian architecture, blurry background, full body, random color vest, short skirt
Paste the following spell in the Negative Prompt field:
paintings, sketches, (worst quality:2), (low quality:2), (normal quality:2), lowres, normal quality, ((monochrome)), ((grayscale)), skin spots, acnes, skin blemishes, age spot, glans, nude
Step 2 AI drawing settings
Scroll down and you will see a lot of AI drawing settings, you can adjust the size of the output picture and make the details of AI drawing. Friends who are playing for the first time can follow the following settings:
Sampling method: DPM++ SDE Karras
Sampling steps (the number of sampling steps, the more the photos, the finer the photos): 30
Restore faces: Be sure to check this option
Width (picture height): initially set to 512
Height (picture width): initially set to 512
(After playing a few times and familiarizing yourself with it, you can decide the size of the image output by yourself)
Batch count (number of output images): 4 (that is, generate 4 images at a time)
Batch size (how many batches of images are output each time): 1 (that is, only one batch of images is generated at a time)
Leave the other options alone. Please note that Restore faces must be checked.
Step 3
After entering the spell and setting the AI drawing options, just click “Generate” to instruct your cloud computer to generate a picture.
Please scroll to the bottom of the screen, you can see that the AI image is being calculated, please wait for the progress bar to reach 100%.
Step 4 output finished product
When you’re done, you’ll be able to see the beautiful girl image generated by the spell you entered. In the settings, the editor sets 4 output images, click to enlarge them, and press “Save” below to save the images directly, or press “ZIP” to download all the images as ZIP files at once (mobile phone users recommend one by one Zhang download is more convenient).
Saved photos, iPhone/iPad users can open the “File” app to open.
We do not recommend using pictures for purposes other than personal viewing and private sharing, so as to avoid any legal risks of claims. Please be responsible for distributing pictures made by AI.
———The most basic teaching is over, the following is a more in-depth AI drawing teaching——
In-depth AI drawing and more ways to play
The most basic Stable Diffusion is introduced above to make AI pictures. Next, the editor will introduce two easy-to-follow gameplays for beginners, and let everyone try to make AI drawings with more freedom.
1) Change the spell yourself
You can notice that the mantras provided by the editor at the beginning are actually all in simple English. You can enter your favorite English keywords (the format is to enter the keyword followed by a “,” comma), and the generated AI image will do what you want.
For example, the editor wants to raise his hands in the image of AI drawing. Just try to enter the following command (simply added to the Prompt spell just now)
arms up, arms behind back,hands behind
The biggest difference between the output photo and the previous one is that the position of the AI characters in the picture will be controlled by the user. A private server is used in this tutorial, so you can freely adjust the AI settings to get the character image you want. The following is the new photo output after adding the new command.
2) Join the profile to play face change, Cosplay change shirt
After you have finished playing different characters, the next thing you want to play is to change the clothes and faces of the AI characters, and make AI character images of different styles. The Stable Diffusion introduced this time supports a character setting file called “LORA” format. The setting file in the “LORA” format can be used to change the clothes of your AI characters, and even change the face of a certain celebrity or Internet celebrity, or shape the AI characters into different figures and physiques.
Step 1 Choose the dressing method you want to play
Beginners don’t know how to play, just watch how others play. Please go to this Civitai website that collects many useful “LORA configuration files”:
URL: http://www.civitai.com
▲You can see different AI portrait models. Find your favorite image and click to enter;
Step 2
Enter the LORA page, scroll to the discussion area at the bottom of the screen, you can see how players use this LORA configuration file to make different AI images. Randomly click the (i) button in the lower right corner of an image. Here are the spells needed to make this AI character image.
Step 3 Copy spells and settings
Seeing these spells, you can directly use them to draw AI images, so you don’t have to worry about giving instructions yourself. All you have to do is copy and paste the spell into your Stable Diffusion.
Computer browser:
Just click the “Copy Generation Data” button below to copy spells such as Prompt, Negative Prompt and other settings. Please click the “Paste” button (shown in red circle) under the Generate button on the Stable Diffusion screen.
Smartphone, Tablet:
Please copy and paste Prompt and Negative Prompt to the corresponding fields of your Stable Diffusion interface, and change other settings by yourself.
▲ Apply the settings provided by the author. For example, there are 3 settings in this picture that need to be modified:
Sampler: DPM++ 2S a Karras
Sampling: 40
Seed: 3628792468
Step 4 Download the LORA configuration file required for the image
The spells alone are not enough, we also need the LORA profile to make these images. Please open the Stable Diffusion interface, and select “CivitAi” from the top menu. Here is a LORA downloader.
A)
Please refer to what LORA is used in the Prompt spell. Taking this example as an example, the published LORA spells are:
B)
The above 4 lines represent 4 different LORA configuration files used in this example. Now we’re going to download them all to your server.
Content type: LORA
Search by term?: Please tick this item
Search term (keyword): Please enter the keyword in the LORA mantra, such as “Miko”;
(Tips: Don’t type yaeMikoRealistic, only part of it)
Press the “Get List” button to start the search;
C)
Click the “Model” menu below to find the “Yae Miko | Realistic Genshin LORA” item used this time from the search results;
D)
Select “YaeMiko-Mixed- Yae Miko | Realistic Genshin LORA” in the “Version” item next to it;
E)
Select “Model Filename” as “yaeMikoRealistic_yaemikoMixed.safetensors”;
F)
Important step: Uncheck the option “Save LORA to additional-networks” that has been checked in the screen, that is, leave it blank;
G)
Click “4th Download Model” to download this AI model to your server. (This screen will not display any response, if necessary, you can return to the Stable Diffusion screen to see the newly added log)
Step 5
By analogy, the editor downloaded the 4 LORAs used in this example to the Stable Diffusion server. The following is the screen when the author downloads these 4 LORA, for your reference.
Step 6 Change the LORA version in the spell by yourself
The LORA downloaded this time may be different from the version called by our copy and paste netizen’s incantation. So you need to check it closely again (press “txt2img” on the upper menu of the Stable Diffusion screen to return to the AI painting screen), and change the spell.
For example, you downloaded 4 different LORA this time, you need to change it to the correct version number. For example, change iu to v35 (shown in red). Also this time we use “koreanDollLikeness“Substitute for “koreanDollLikenesss“, so you also need to change the name and version number of LORA in the spell (shown in red).
The red text shows the revised version number/LORA name:
Step 7 Generate AI image
Now you can use the spell you just copied and press the “Generate” button to generate the image.
In this way, you can generate some AI character images that have changed their clothes, and the settings follow the Civitai web page settings.
Play more resistant and run for a long time
The free version of Google Colab has certain usage restrictions. We have tried to shut down the virtual computer after about 10 hours of operation, and all the work of downloading LORA has to be done again. If you need to use it for a long time, the personal Stable Diffusion server settings will not disappear, you can subscribe to its paid version “Calob Pro”, the lowest fee is HK$77.80 per month.
▲After the virtual computer is turned on for about 10 hours, it reaches the upper limit of continuous startup time and needs to be forced to shut down.If you want to continue using the free version, you can start a new execution phase and re-execute the instruction set
latest videos