Important Prompting Tips - Stable Diffusion, Fooocus, Midjourney and others 

Happy to help! So many little things and I am always finding new things the more I experiment. The association effect is actually a big factor.

Some of these are starting to get better at understanding prompts. Good to know.

Thank you 👍

Absolutely!

Love it! I like doing these because it gets you to think about how the AI handles things.

I think that if you see the engine fighting the results it means that you should turn something off

@@marcelobaus Indeed; the generation of the tires is what I was working to stop.

Excellent. I'm still trying to find a way of getting a newer car without tires as well. That one is pretty tough.

Welcome

I did a little experimenting with multi line prompts and don't think you would get the expected results separating sections of prompt on each line. You will get some of what you might expect but not all. At least what I messed around with. For example I just tried using the ones you just listed and never got a full moon or a wolf. But once I combined on one line it did give me those more often.

@@KLEEBZTECH ah okay, i havent this issue and i use this Multi-line method only too organized my prompts because you know sometimes if you want something good you make a mess with many prompts and i have often this problem that i dont understand what prompts mess up my image and with this method i can play around which word make a clown out of me 😄

Sounds like Openpose. Fooocus does not support that yet but I think it may in the future. I do know there is a fork of Fooocus that supports it but that fork last I checked was way out of date. I have used with other tools and have mixed feelings on how well it works. Works better with SD 1.5 and not so good with SDXL. But would be nice to see it added.

8K just refers to 8k resolution like 4k. Just terms used to try to get higher detail usually. Using different camera terms can have an impact on your prompts. I am not sure I if I tried tireless yet but will have to give it a shot tomorrow. 👍

If you mean use other Stable Diffusion tools like A1111 or ComfyUI you will get different results. Fooocus has some backend tweaks. Fooocus is much easier to use but the others do have some benefits in some cases. I use several myself.

@@KLEEBZTECH The fact is that I made an Ai influencer created in foocus, but I want to use loras 1.5 but they don't work in foocus. If I use the same model in stable diffusion, the realisticstock, would it have different results for the influencer?

I might not be the same but probably close. Hard to say without trying. Just to mention you can use 1.5 LoRAs if you use a 1.5 model as a refiner in Fooocus but not sure how much that will effect your influencer. But I find that any LoRA is going to have an impact on the results as well.

I have not... But the challenge was to remove tires, not clothing... 😂

So it does not check off the box when you click on it? Weird. Have you tried a different browser? Is it actually applying the styles if you do click on them? Have you checked the log to see?

@@KLEEBZTECH I don't know how to open in a different browser. It does apply the styles. I checked the log and it showes the styles selected.TIA

@@earlsargent105 Do you have issues with any of the other checkboxes in the interface? If not then might have nothing to do with the browser. But if you have a different browser installed you would just need to enter the same address in the address bar and it will open in that browser.

@@KLEEBZTECH I have the same issues with all the checkboxes in edge but when I oped opera they all work. The styles do not show up in opera but I checked the symbol and it is rotating so I am waiting to see if they show up. They showed up. Now I am checking to see if I can modify the run programs to open opera instead of edge. Thank You for your help.

I wonder if there is something blocking certain things. I just tested in Edge and it all worked for me. I normally use Brave without issue and I also know Chrome should have no issue.

No luck, but by including the word wheels it is also more likely to include wheels.

And that is the type of thinking you need when creating good prompts. Think of circumstances that would have the tires off and prompt for that.

@@KLEEBZTECH I played with this some more today and just had fun with it "automotive manual exploded view of (entire car) brakes, suspension, engine 1950's"

And now we have parts all over the place. lol. Have you tried adding the "Knolling Photography" style to it?

​@@KLEEBZTECHalso when I enable the face swap i get 1006 in the prompt, and the focus get stuck

You will need to follow the instructions for AMD devices. You need to add more to your run.bat file. Will warn that AMD will be pretty slow. See here: github.com/lllyasviel/Fooocus?tab=readme-ov-file#windows-amd-gpus

@@KLEEBZTECH Thank you very much! Its working!

Also i have the same issue with google colab fooocus and its confusing since it is running on their server

As for Colab is it after some time or when you first try? Because what will happen with Colab is when you hit the usage limits it will not allow you to use with a GPU until the limit gets lifted the next day. I will say I don't use often on Colab so not super knowledgeable about using on it.

@@KLEEBZTECH Yeah, its seems that situation is what you desribed. Thank you again!

Are you describing details of the face? If you describe things like eye color and such that will happen. Also use things like casual for example can help. But really the big thing is usually describing facial details will cause that to happen.

🤔

should we?

Are you using any styles when you generate and get them? If you can give me more info I might be able to give you some suggestions.

I appreciate that. The situation is when I am trying different variations of input images. I'll drop an image of a portrait into the Upscale/Variation window, checkmark Vary (Subtle), add some text prompting; usually a sentence followed by some descriptors: "a bi-racial man with blue eyes and a buzz cut dressed in golden full armor, white background, solo, looking at viewer, smile, blue eyes, simple background, black hair, 1boy, white background, upper body, male focus, teeth, dark skin, armor, grin, dark-skinned male, shoulder armor, portrait, pauldrons, realistic, gold armor, golden armor, solid gold breastplate, mixed ethnicity," Presets are on: Realistic, Speed, 1:1, PNG, and the Negative Prompt will have "unrealistic, saturated, high contrast, big nose, painting, drawing, sketch, cartoon, anime, manga, render, CG, 3d, watermark, signature, label" Selected Styles: Fooocus V2, Fooocus Photograph, Fooocus Negative. Base Model, usually one of these checkpoints; juggernautXL_version6Rundiffusion socarealismXL_vae30 sdxlTripleHeadedMonkeys_v10 realisticStockPhoto_v10 devlishphotorealism_sdxl15 Sometimes with or without a Refiner (from the same list of checkpoints) Only LoRA would be SDXL_FILM_PHOTOGRAPHY_STYLE_BetaV0 (weight 0.25) No settings on the advanced tab are changed. I will get ear rings 90% of the time. I will add to the Negative Prompt line of text: "no ear rings, no earrings, no ear-ring," etc Still around 85% of the time I'll get ear rings. I'll try adding to the text prompt: "empty ears, bare ears," etc I'll get ear rings 95% of the time then. I blame the checkpoints, but I do like the other results I get from using them. I think the real solution will be inpainting, but I haven't tried that yet. I'm sure that is something you have covered in your videos, so I'll poke around and look. If you ever do a video on using Oobabooga's text-generation-webui to implement the use of Fooocus I'll be very interested in seeing it!!

@@marshallodom1388 interesting. My first thought was you had the Masterpiece style enabled and that includes the word earring. I'm not seeing anything that stands out that would explain it but I'll think about it. Does it do it when just generating as well? Or just under the vary/upscale section? As for negative prompt, just put earrings, no need to add the word no. Negative prompts only have minimal effect but also you can add extra weight (earrings:1.5) for example.

@@marshallodom1388 I wonder if using the word gold is part of the issue. That gets into the association effect. Often that's the sort of thing that causes these things that I find.