<h2 id="cinematix-inpainting-(new)-wsqqe6yuq">Cinematix Inpainting (NEW)</h2>In Forge, set XL as your model type and select cinematix.inpainting.safetensors as your model. Go to Img2img tab to Generation/Inpaint. Load your image and mask the area you want to change. Set a simple prompt describing the changes. Generate. Change the Denoising strength if you want a stronger effect. You should get nice outputs right away.Experiment with Inpaint area Whole picture/Only masked. If using Whole picture, set the image size to your input image aspect ratio.Also worth checking out:Cinematix CosXL EDIT is realistic SDXL Pix2Pix model for editing. It uses a different workflow and you can find it here:<a target="_blank" rel="ugc" href="https://civarchive.com/models/396070/cinematix-cosxl-edit">https://civarchive.com/models/396070/cinematix-cosxl-edit</a><h2 id="cinematix-l8-inpainting-q90odirw4">Cinematix L8 Inpainting</h2><ol><li>Use CFG 1.5-4 for 10-20 steps, Euler A, DPM++2Sa + SGMUniform schedule</li><li>Created for Flux experiments, check also <a target="_blank" rel="ugc" href="https://civarchive.com/models/106055?modelVersionId=263752">Photomatix SD Inpainting model</a> for fast sketches</li></ol><h2 id="cinematix-l8-5oliyopnl">Cinematix L8</h2>Performs better in upscaling, needs more steps. Cinematic photo style.<ol><li>Lightning Model, Use CFG 1.5 for 10-12 steps</li><li>For more steps, adjust CFG proportionally (~ CFG 2 for 24st., CFG 2.5 for 36st)</li><li>Works best with Euler A + SGMUniform schedule</li></ol><h2 id="cinematix-l4-juqxe4ip4">Cinematix L4</h2>Scenes should be better defined when using cinematic styles.<ol><li>Lightning Model, Use CFG 1.5-2</li><li>Steps 7-12 for Euler A, DPM++ 2S a, and similar samplers for a nice single-pass image (should generate a nice picture with minimum artifacts)</li><li>More steps can be used for the LCM sampler</li><li>Recommended 'Euler A SGMUniform' sampler (Forge)</li><li>For best details, experiment with a 2-pass workflow (in A1111, use Hires fix tweaks and LCM in the second pass)</li></ol><h2 id="cinematix-lightning-dguuq3mei">Cinematix Lightning</h2><ol><li>Use CFG 1.5</li><li>steps 5-12 in a single pass for a nice picture</li><li>LCM or Euler A SGMUniform sampler (standard samplers can/will output strange colors or artifacts)</li><li>2-pass can use 3-4 steps + 3+ steps in Hire fix, Denoising strength around 0.5</li><li>Note that SDXL-Lightning performs best in resolutions around 1024</li></ol>More about the workflows in <a target="_blank" rel="ugc" href="https://sandner.art/sdxl-lightning-how-to-use-it-for-the-best-details/">this article</a> on [sandner.art].<h2 id="cinematix-sdxl-v1-kbthkm5yo">Cinematix SDXL v1</h2>Cinematix is a very fast realistic SDXL model. Read more details in an extensive <a target="_blank" rel="ugc" href="https://sandner.art/cinematix-sdxl-expert-tips-and-advanced-techniques-for-fine-tuned-stable-diffusion-models/">article on my pages</a>.<ol><li>You may use standard samplers with lower steps 5+ (but you should go higher), recommended 12 steps for most scenes</li><li>Samplers: DPM++ 2S a, LMS Karras, Euler a (LCM not performing well in low steps)</li><li>You may use Photomatix and Animatix experimental styles for great results. <a target="_blank" rel="ugc" href="https://github.com/sandner-art/Photomatix/tree/main/PX-Styles">Download them here</a>. Read the instructions first. There are classic SDXL styles for your reference tests too.</li><li>You may use higher CFG (4-7+), instead of the usual LCM CFG 1.5.</li><li>An extensive negative prompt is NOT needed. Actually, in most cases, the negative prompt is not needed at all. Check the gallery for examples.</li><li>You may be very brief when prompting. Examples in the gallery are for comparison with my previous checkpoints and posts.</li></ol>If you create something interesting with it, post a review.Cinematix was created by merging and experimenting with these models:StabilityAI <a target="_blank" rel="ugc" href="https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0">SDXL</a><a target="_blank" rel="ugc" href="https://civarchive.com/models/238116?modelVersionId=275681">OpenDalle</a><a target="_blank" rel="ugc" href="https://civarchive.com/models/241445/hephaistos-nextgen-dpo-swfandnsfw">Hephaistos</a><a target="_blank" rel="ugc" href="https://civarchive.com/models/254191/raffaelloxl-real-people-photography">RaffaelloXL</a>SDXL-Lightning <a target="_blank" rel="ugc" href="https://huggingface.co/ByteDance/SDXL-Lightning">https://huggingface.co/ByteDance/SDXL-Lightning</a>, <a target="_blank" rel="ugc" href="https://arxiv.org/abs/2402.13929">https://arxiv.org/abs/2402.13929</a>L8:<a target="_blank" rel="ugc" href="https://civarchive.com/models/299933/halcyon-sdxl-photorealism">HalcyonSDXL</a><a target="_blank" rel="ugc" href="https://civarchive.com/models/161068?modelVersionId=275491">NewReality</a>

Cinematix

cinematix_l8.safetensors

v1-inpainting

L8-inpainting

v1.4 Lightning

Cinematix L8Performs better in upscaling, needs more steps. Cinematic photo style.<ol><li>Lightning Model, Use CFG 1.5 for 10-12 steps</li><li>For more steps, adjust CFG proportionally (~ CFG 2 for 24st., CFG 2.5 for 36st)</li><li>Works best with Euler A + SGMUniform schedule</li></ol>

Files with SHA256: f6f57c8bff01e7d00c8f326443c41cba3b90654a0056580195b9de2998e0b69b