Reiko's Guide on AI Generated Images
Where do i start?
The easiest way to start is just going to Pixai or Tensor, login and start writing prompts, using other artworks as references.
| Tensor | Pixai |
|---|---|
![]() |
![]() |
My AI generations look horrible and just generates nightmares [Settings]
Relax, most of the time there's something wrong with the settings, or that your prompts are not optimized/making the AI confused on what you're trying to generate.
| 1 | 2 |
|---|---|
![]() |
![]() |
Above is a default setting i like to use on Tensor, I recommend CFG Scale on 3-7 anything above or below that, is going to give you generally bad gens, and for sample and scheduler, Euler A and Normal always works, another combo that usually works is DPMPP 2M+Karras, but don't expect it to work all the time.
| 1 | 2 | 3 |
|---|---|---|
![]() |
![]() |
![]() |
Above is a default setting i use on pixai, same thing as Tensor, reccommended CFG Scale is 3-7, scheduler for Pixai should just be Euler A with normal scheduler, because anything else is more expensive, and in terms of credits, pixai is not as budget friendly as Tensor, even if they're both free to use.
My gens are being blocked/I can't see the image i just generated [NSFW Toggle]
This is mainly a problem with using Tensor on the mobile app, instead of using it on your web browser, because the app blocks all NSFW generations/features, if you're already on your browser, open your settings, and turn on the Mature Content switch, that should fix it.

How do i start making AI Generated images? [Prompt]
On Tensor and Pixai you're going to need to learn prompts, but most of the time you can just use the remix button from posts you see and just start from there, for now i'm just going to give you something simple and default, and rarely breaks. (I'm using the settings i showed above)
1girl, small breasts, black hair, long hair, twintails, purple eyes, mole under mouth, glasses, collared shirt, pleated skirt, smile, sitting on couch, living room, looking at viewer, BREAK hyper-detailed, high contrast, depth of field, ray tracing, best lighting, cinematic composition, beautiful face, beautiful eyes, sharp focus, (masterpiece:1.2), (best quality:1.2), (very aesthetic:1.2), (absurdres:1.2),perfect anatomy, detailed shading, High details, light particles
negative prompt: modern, recent, old, oldest, cartoon, graphic, text, painting, crayon, graphite, abstract, glitch, deformed, mutated, ugly, disfigured, long body, lowres, bad anatomy, bad hands, missing fingers, extra digits, fewer digits, cropped, very displeasing, (worst quality, bad quality:1.2), bad anatomy, sketch, jpeg artifacts, signature, watermark, username, simple background, conjoined, bad ai-generated,EasyNegative,BadDream, bad anatomy, poorly drawn, bad anatomy, wrong anatomy, lowres, text, error, cropped, worst quality, low quality, out of frame,blurry,username, watermark, signature, censored
| Tensor With Quality Prompt | Tensor Without Quality Prompt | Pixai Ver. With Quality Prompt | Pixai Ver. Without Quality Prompt |
|---|---|---|---|
![]() |
![]() |
![]() |
![]() |
So, explanation, you can see that my personal prompt is divided into something like "Subject, Body Details, Clothes/Accessories, Actions/Position, Objects/Backgrounds" But then you also have the BREAK into some high quality stuff, that's what i like to call quality prompt, it's most notable on Pixai compared to Tensor, and you also have negative prompts, it's useful for when you wanna make something but when the AI generated something you don't want, just add that in the neg prompt.
What model should i use? There's so many choices! [Models & LoRAs]
Anything honestly, a lot of them are really good, but here are some that i often use and recommend or popularly used by others, these are the ones i use on tensor since i predominantly use that nowadays,
| 1 | 2 |
|---|---|
AniCoreXL IllustriousV5.1 ![]() |
Nova Orange XL Crody ![]() |
Hassaku XL (Illustrious) ![]() |
Amanatsu ![]() |
Aüngir ![]() |
JunkJuice Velvet ![]() |
WAI-ANI-NSFW-Illustrious ![]() |
Paruparu Illustrious ![]() |
What are LoRAs? What LoRAs should be used? I wanna use a style from a certain artist! [Models & LoRAs]
LoRAs are small adapter models that you use to finetune your image gen even more to what you want, so should you use them? yes, if it helps you make the image to be better/looks more to what you want. Want to make an AI Generated image to look like a certain artist? Find it on the LoRA search, 90% of the time someone already made it and you can just plop it on there and add all the trigger words, and voila, now it looks like an imitation of that certain artist
| 1 | 2 |
|---|---|
Ratatatat74 style on Tensor (Nova Orange + LoRA) ![]() |
Ratatatat74 style on Pixai ![]() |
AIイラストおじさん on Tensor (AniCore + LoRA) ![]() |
Anime Pastel Colors V2 on Tensor (AniCore + LoRA) ![]() |
Sharp Semi Realistic GNRA on Pixai (Amanatsu + LoRA) ![]() |
Shiny High Definition GNRA on Pixai (Amanatsu + LoRA) ![]() |
AI Ethics & Moral Obligations
This is my opinion on what you should and shouldn't do with your AI Generated works, me personally, i never post it outside the website i generated it in or other AI based websites (JanitorAI or ChubAI), i never post it on a space where someone could mistake it as real art, e.g Twitter, Instagram, Pixiv, etc. But if you do (You shouldn't) Always ALWAYS mark your work as AI Generated, don't make it hidden or pretend it's not AI Generated, because as an artist myself (I draw) seeing AI Generated Art on my timeline that doesn't explicitly say it's not AI Generated hurts me and all other artists. And I will hurt you in your dreams if you do that.























