<hr />DisclaimerFinancial support is very much appreciated, but please do so by donating to my <a target="_blank" rel="ugc" href="https://ko-fi.com/aicharacters">Ko-Fi</a> instead of sending me Buzz, as I currently have no uses for CivitAI Buzz!<hr />DescriptionAn SDXL 1.0 LoRa made by AI_Characters for producing images with an artstyle similar to the artworks produced by Jakub Rozalski - creator of the fictional alternate history Dieselpunk with mechs setting 1920+ in which also the boardgame Scythe and PC real-time tactics game Iron Harvest are set in.<hr />Recommended generation parametersWeight: 1.0 (adjust down as needed when used in conjunction with other LoRa's)Model: <a target="_blank" rel="ugc" href="https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/blob/main/sd_xl_base_1.0_0.9vae.safetensors">SDXL 1.0 with 0.9 vae</a>Steps/Sampler/CFG/Resolution: 25 steps | DPM++ 2M Karras | CFG 7 | 1536x1536'Kohya Deep Shrink' settings: 4 blocks | 2 downscale | 0 start | 0.35 end | skip = truePositive prompt: high-definition 8kNegative prompt: text, blur, low-resolution, low-quality, grainy, glossy, poorly drawn, amateur, poorly shot, compression artifacts'CLIPTextEncodeSDXL' ComfyUI node resolution: 4096x4096<hr />Trigger words (must be included in the prompt for the LoRa to take effect!)<ul><li>'painting in jrst artstyle'</li></ul><hr />Further usage notes/<hr />Usage permissionsYou may not sell this model, merges containing it, or images generated with it.Credit me when using this model in a merge or hosting it on a generation service.<hr />TL:DR Training Workflow<ol><li>Train full finetune with a dataset of 20-100 images using <a target="_blank" rel="ugc" href="https://www.dropbox.com/scl/fi/r6r7894spm6el0togcc4u/SDXL1.0_Essenz-series-by-AI_Characters_Style_HighQualityHeavilyPostProcessedPhotography-v1.1-Hal-_20231118-093204.json?rlkey=xm9u9b45me7slrtsm03q02qv9&amp;dl=1">this config</a> on a <a target="_blank" rel="ugc" href="http://Vast.AI">Vast.AI</a> rented A100 80gb GPU</li><li>Extract LoRa at 512/512 dimensions using <a target="_blank" rel="ugc" href="https://imgur.com/a/XRare3P">these settings</a></li><li>Resize LoRa to 16 ranks using <a target="_blank" rel="ugc" href="https://imgur.com/a/PZiGEtt">these settings</a></li></ol><hr />Changelogv1.2 [FINAL]<ul><li>complete rework; higher epoch count, trained with EMA, reworked captions, reworked dataset preparation, small trigger words change - overall much better likeness and quality at the cost of some slight bias and overtraining increase - final version for now</li></ul>v1.0<ul><li>Initial release</li></ul><hr />

Files with SHA256: 8c67ea86d9f5b88d64421e019e3f581aa5ad9709b379eae7fddd46b55df2dacd