You can find out more about realism settings and workflows here: <a target="_blank" rel="ugc" href="https://civarchive.com/articles/32409/realism-how-to-achieve-it-a-generation-guide-workflow-examples-included">https://civarchive.com/articles/32409/realism-how-to-achieve-it-a-generation-guide-workflow-examples-included</a> I decided to start from scratch and took two days to build a workflow to block merge, then merge, then inject💉, then finetune. and then failed... <h3 id="v3.1">V3.1</h3>Welp.... after exhaustive testing and a rather priey API call, I think I've delivered a hell of a model. Read all about the journey and the testing Here: <a target="_blank" rel="ugc" href="https://civarchive.com/articles/35724/model-same-face-syndrome-how-to-fix-it-without-retraining-test-sheets-included">https://civarchive.com/articles/35724/model-same-face-syndrome-how-to-fix-it-without-retraining-test-sheets-included</a><h3 id="v3.0">V3.0</h3>I tired merges, images, messing with block swaps, epochs, etc. I finmally used a mix of all of them.<ul><li>slightly more responsive to NSFW (not my goal. There are LoRAs for that)</li><li>better prompt adherence.</li><li>Better asthetics (I think)</li></ul>Let's hope thie version measures up.<h3 id="v2.0-(nsfw-version)">V2.0 (NSFW version)</h3>Does NSFW exceedingly well. Beats most models here, but it does not measure up to V1.0 realism<h3 id="v1.0">V1.0</h3><ul><li>Trained on somethin like 3600 images for two weeks straight.</li><li>semantic tuned to get that SDXL look</li></ul> All images I uploaded are Euler/ simple, 10-12 steps, no LoRAs, using my Pro workflow here: <a target="_blank" rel="ugc" href="https://civarchive.com/models/2726952/krea-2-pro-grade-w-image-edit-style-transfer-moodboard-controlnet-multi-reference-sam3-detailers-and-low-vram-options">https://civarchive.com/models/2726952/krea-2-pro-grade-w-image-edit-style-transfer-moodboard-controlnet-multi-reference-sam3-detailers-and-low-vram-options</a><hr />Instagram: <a target="_blank" rel="ugc" href="https://www.instagram.com/synth.studio.models/">https://www.instagram.com/synth.studio.models/</a>Buy me a☕ <a target="_blank" rel="ugc" href="https://ko-fi.com/lonecatone">https://ko-fi.com/lonecatone</a>This represents many hours of work that I give away for free. If you enjoy it, Please make sure to 👍like and post to my models &amp; workflows. That is how you support me!! Also, feel free to ⚡tip 😉

Krealism

FP8 VersionV3.1 fixes a face lock issue

krealism_v31TurboFP8.safetensors

V3.1 Turbo FP8

V3.1 Turbo nvfp4

V3.1 Turbo Int8_convrot

V3.1 Turbo bf16

v1.0 Turbo nvfp4

V2.0 Turbo NSFW FP8

V2.0 Turbo NSFW nvfp4

V2.0 Turbo  int8_convrot

V2.0 NSFW Turbo_bf16

v1.0 Turbo FP8

v1.0_turbo_q8_0.gguf

v1.0 Turbo int8_convrot

v1.0 Turbo bf16

Files with SHA256: fadea114dfc0f9a1a9583642026b4433e603c345379adf07810db118704c4f31