No Lora’s are required I got great results just by prompt engineering it’s fantastic 🤩
Lora’s are useless as they ruin the quality of your images unlesss you crank that resolution up high.
This system writes your idea enhanced in photography tone, it’s really awesome u can modify the structure as needed u can use anything like movie, videogame, art like Greg rutkowski and more. Enjoy btw u can use google ai edge gallery app as they have the newest Gemma llm its better even on 2B version. Enjoy and mod to your liking🤪🤓👍🏽😉👀
Also put this in clip L: Grimdark Fantasy rendering aesthetic style, aesthetic_10
Clip G: Unreal engine 4 aesthetic rendering style
I’m a kinda rookie prompt engineer but have discovered new way to get llm/vlm to speak in different styles which gives different results it’s awesome. Try the style byte tool and some extras the Omni style has an pg-13 and an unedited version. Enjoy the Omni version is all the linguistic filters into one powerful system Prompt😎🤓😉
Description
I must make this clear this has nothing to do with midjourney I was inspired by there images and wanted to try to get midjourney like images in stable diffusion. No Lora’s are needed for this. The style tool zip also have xtra tools that I’ve created enjoy😉
If you use the Omni version only zimage understands it as other models will create deformation mess. Flux.1 krea also understands.
* Style Transfer via Image Prompting: Paste the style tool prompt, upload an image, copy the resulting style, and then erase the original tool prompt.
* Style Injection: Paste the copied style (as a byte) into the system prompt box.
* Style Enforcement: Any subsequent text input will be generated by the VLM/LLM in the newly injected style.
You need to use an vlm and llm with thinking reasoning features or you won’t get the best results.