Trained on a cleaner dataset. It should now have a more consistent style.










Hello everyone! Lately in the comments I have often seen people complaining that my LoRA are similar to each other or are not representative of the style they should represent. My models are not trained on material from other artists, and it is therefore difficult to explain what the actual output should be. Since preparing the material necessary for the publication of a LoRA usually takes me a day's work, I have therefore decided not to post all my LoRAs on CivitAI anymore, but from today I will only post those that I consider interesting enough for the audience of this site. If you are interested in downloading the other LoRA as well, you will find them all for FREE (and they'll always be) on my Patreon page (subscribers will have early access to each of these). Thank you all for your support!
This is a 80s fantasy movie based LoRA. It's trained on movie posters, so I advise using (text,title,watermark,signature) in the negative prompt to avoid text. This is an experimental LoRA, so the quality of the output can change a lot based on the weight and the prompt.
All the images in the examples have been created using my custom model (https://civitai.com/models/24149) and this LoRA only. All the examples have been created using the hires-fix or by creating the base 512x768 image, upscaling it to 1.5 and the applying the same prompt again as img2img
As the LoRA is trained on real pictures or drawings, use a config between 0.5-0.7 if you want to keep your original art style.
The LoRA works also on ???? (check last examples).
There are now 2 different versions:
v1: trained on 80s movie posters scraped from the web
v2: trained on the same dataset, but this time cleaned from text and colored borders. The output of this version is cleaner and more defined.
Learn how to create pictures like mine with my step-by-step tutorials:
Check out my Patreon for exclusive content:
https://www.patreon.com/Inzaniak
Do you want a custom LoRA or picture? Check out my commission page:
If you want to support my work for free, you can also check out my music/art here:
Trained on a cleaner dataset. It should now have a more consistent style.
1. 재게시된 모델의 권리는 원 제작자에게 있습니다.
2. 모델 원작자가 모델을 인증받으려면 공식 채널을 통해 SeaArt.AI 직원에게 문의하세요. 저희는 모든 창작자의 권리를 보호하기 위해 노력합니다. 인증하러 이동
