Update: Following a request, i reposted the old versions "IXL Next" and "IXL V0" here as previous versions of Reborn.Usage<ul><li>Steps 25-30 should be enough, 40 doesn't hurts :D</li><li>CFG: i tested with 5, going lower may be fine</li><li>Scheduler Sampler: Euler A and DPM++ 2M should be fine</li><li>Clip Skip: the usual 2</li><li>Adetailer: recommended :D</li><li>Negative prompt: IllusN3g, deep_negative_pony (for a more authentic feeling :D)</li></ul>Adetailer trick: To avoid the "box effect", what i usually do is set this in the inpaint section:<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/b00e5c88-2006-4b2a-8b1b-999e592e0a4e/width=525/b00e5c88-2006-4b2a-8b1b-999e592e0a4e.jpeg" />Long Story™One thing has been frustrating me for a long time. Neither the Next or V0 version of BancinIXL felt close enough to BancinXL. So, i deleted those got back to work.Here, i started back from scratch to try and get a closer feeling, using two different approach using AnBan 1.0:<ul><li>Using supermerger with "Add Difference" in "Train difference" using BancinXL-Pony (leveraging xyz plot, i got that 0.5 was the max i could incorporate this way)</li><li>Directly finetuning the checkpoint using the dataset (200 pictures) i previously used for a BancinXL LoRA (the method was the same used for HoJ, but since the dataset was small, it got a strong bias)</li></ul>This left me with two differents checkpoints that i merged together with an intermediate step for each (closer to merge, closer to finetune), carefully selecting the block pairs that i liked.<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/b988016c-6f1c-4b23-8317-f792f21b174d/width=525/b988016c-6f1c-4b23-8317-f792f21b174d.jpeg" />Feel free to experiment, this is still not a "Pony like illustrious", but it should feel more western/comic style that AlterBan or Quirk 😅

BancinIXL - Reborn

Reupload of IXL V0 as an old version for Reborn

bancinixlReborn_ixlV0.safetensors

IXL V0

IXL Next

IXL Reborn

V3!As I am working on a new Pony model for fun, I did a quick side-quest and created a V3 for BancinXL! Based on Juice and V2 with some LoRA and MeichiDarkmix checkpoints, i'll do a longer story later 😉 (PS: sorry for the messy showcase pictures, it's been a while since i generated Pony stuff and i forgot how much it is difficult to get good hands with it)TL;DR<ul><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/875403?modelVersionId=1088540">V2</a> is the main Pony version, including the "classics" LoRA (InCase/Vixon/Expressiveh/Disney)</li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/875403?modelVersionId=1108036">Juice</a> is variant of V2, more shiny, more crazy</li></ul>Juice!<a target="_blank" rel="ugc" href="https://civarchive.com/models/547592?modelVersionId=663388">PsychoNex</a> is like a hard liquor, tasty but drink it pure and you'll go blind ;-)I have told in many posts that i love these models, but sometimes, working with them is not yielding what i wanted (but maybe, what i deserved :D).The Juice version of BancinXL is simply a 50/50 merge mix of PsychoNex and V2.Not intended to be the main version, but still a whole lot of fun for more chaos :DGo check <a target="_blank" rel="ugc" href="https://civarchive.com/user/skumerz">Skumerz</a> work!V2Rework of the V1, with a similar approach, with just a few more LoRA to give it some oomph.Still based on <a target="_blank" rel="ugc" href="https://civarchive.com/models/257749/pony-diffusion-v6-xl?modelVersionId=290640">Pony XL v6</a>, before jumping on the Illustrious/NAI XL trend :DRecipe:BancinBase (using SuperMerger):<ul><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/212532/all-disney-princess-xl-lora-model-from-ralph-breaks-the-internet?modelVersionId=239420">All Disney Princesses V1</a> (1.0) [Tried the Pony version, V1 still the better fit]</li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/341353/expressiveh-hentai-lora-style?modelVersionId=382152">ExpressiveH</a> (0.65) ❤</li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/300005/incase-style-ponyxl?modelVersionId=436219">InCase Style [PonyXL] V3</a> (0.55) ↑ 0.05</li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/888231?modelVersionId=1029540">Vixon's Pony Styles Gothic Neon V2</a> (0.35) ↓ 0.1</li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/11161?modelVersionId=839103">Cutesexyrobutts Style PDXL V3</a> (0.25) 🆕 [After dimension rework from 16 to 32]</li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/333041">Vintage 1990s Anime SDXL LoRA (Style) Pony V3</a> (0.2) 🆕</li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/264290?modelVersionId=333587">Style for PonyXL V6 - Smooth 2</a> (0.15) 🆕</li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/122359">Detail Tweaker XL - V1</a> (1.2) 🆕 [After dimension rework from 16 to 32]</li></ul>BancinTemp (using Checkpoint Merger):<ul><li>0.8 x BancinBase + 0.2 x BancinXL V1</li><li>Baked in VAE: <a target="_blank" rel="ugc" href="https://civarchive.com/models/199948/flatpiecexl-vae">FlatPieceXL</a> (switch from SharpSpectrum, still good colors, just less strong)</li></ul>BancinXL_V2 (using SuperMerger):<ul><li>BancinTemp + (0.4) x <a target="_blank" rel="ugc" href="https://civarchive.com/models/711522">Detailed/Perfect Hands</a></li></ul>Usage:I have added a step/sampler comparison to help choose, but the model works good with the classics without any issues. Euler a give more smooth picture while DPM++ 2M give more crisp results.For Euler a, 40 steps give a better looking result, while only 20 steps are enough for DPM++ 2M.DPM++ 2M SDE change the composition a bit but is nice for experiments.I tend to use CFG between 4 and 6 to give more or less liberty to the generation, but that's a personal preference.Regarding prompts, most are done with the classic Pony scores stuff, here is what i do for reference:<pre><code>Expressiveh, (optional LoRA trigger words), masterpiece, best quality, PonyScores7, &lt;1girl|1boy|1other&gt;, (character name as recognized by Pony), (body description), (scene description), (background and lighting)</code></pre>For negative prompt, except for specific, it's always this:<pre><code>PonyScoresNeg, (worst quality, low quality:1.4), (deformed, distorted, disfigured:1.3), poorly drawn, bad anatomy, wrong anatomy, extra limb, missing limb, floating limbs, bad hands, missing fingers, extra fingers, fused characters, muscular female, deep_negative_pony</code></pre>Embeddings used (real SDXL TI, SD1.5 TI are useless unless you hack stuff in ComfyUI, yes, i am looking at you EasyNegative ^^;):<ul><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/831971">Deep Negative Pony</a></li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/649653?modelVersionId=730006">PonyScores7</a></li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/649653?modelVersionId=727303">PonyScoresNeg</a></li></ul>Personal trick:I personally prefer the crispiness of DPM++ 2M for the overall picture, but faces are generally better with Euler a.To mix the best of both, I generate using DPM++ 2M with 40 steps and when <a target="_blank" rel="ugc" href="https://github.com/Bing-su/adetailer">ADetailer</a> kicks in, it is configured to switch only the sampler for the face to Euler a (will keeping steps, prompts, clip skip and so on).Simple and does the trick, didn't want to bother with also choosing CFG/Steps/etc...in the extension. :DV1So. A very simple merge to integrate the LoRAs the amazing creator <a target="_blank" rel="ugc" href="https://civarchive.com/user/Bancin">Bancin </a>use in almost all his/her pictures ^^; (GO CHECK HIS/HER WORKS!)Made to allow trying his "style" on my GPU poor setup without overloading it too much.Base model: <a target="_blank" rel="ugc" href="https://civarchive.com/models/257749/pony-diffusion-v6-xl?modelVersionId=290640">Pony XL v6</a>LoRAs and weights:<ul><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/212532/all-disney-princess-xl-lora-model-from-ralph-breaks-the-internet?modelVersionId=239420">All Disney Princesses V1</a> (1.0)</li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/341353/expressiveh-hentai-lora-style?modelVersionId=382152">ExpressiveH</a> (0.65)</li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/352581/vixons-pony-styles?modelVersionId=398847">Vixon's Pony Styles Gothic Neon</a> (0.45)</li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/300005/incase-style-ponyxl?modelVersionId=436219">InCase Style [PonyXL] V3</a> (0.5)</li></ul>Update: Wow, scrolled the Disney Princesses Gallery and discovered everyone and more are using approximately the same mix of lora ha ha ha. Lots of VRAM consumed on merging those with Pony each time instead of using a pre-baked one ^^;

Files with SHA256: 54a0321f31e9f2981d0f044b333e8ff365d2bb4b5e87fa18a1075f6855adf773