<a target="_blank" rel="ugc" href="https://civarchive.com/models/2626223/narnima-turbo-all-in-one-anima-v1-checkpoint?modelVersionId=2948543">nArnima </a>is strongly influenced by my own preferences, so i wanted a new "base" ANIMA to be able to test LoRA and style, while having its own default style baked in. And while i have a V2 of nArnima-Turbo ready, i still made a new base that I felt like making available in advance during my break.Update:<ul><li>The pruned version is the standard DiT part, which need the TE and VAE</li><li>The full version is the All-In-One, Turbo v1.5 included version</li></ul>Usage:<ul><li>Resolution: With both ANIMA 1.0 and my own training leveraging high resolution for the dataset, pictures can be generated directly at 960x1440 px with good quality</li><li>ER SDE / Beta or whatever you fancy with base ANIMA</li><li>32 steps and CFG 4</li><li>Flow shift 3</li><li>Classic positive: masterpiece, best quality, score_9, newest, detailed</li><li>Classic negative: worst quality, low quality, lowres, score_1, score_2, score_3, anatomical nonsense, watermark, text</li></ul>NB: There is a bias i failed to take into account, characters will probably have ear piercing. Add this in the negative if needed. But why is that?Training storyThis model is the result of an experiment around making a character LoRA with baked in style and allowing the style be used without the character if the tag is missing.... and you guessed it!My OC Liadora is baked in this model. No need for an additional LoRA, if you tag:<ul><li><code>Liadora, red hair, green eyes, black horns, eyeliner, mascara, red lips, curvy, narrow waist, (liadoraoutfit, short dress, pleated skirt, elbow gloves, thighhigh boots)</code></li><li><code>Liadora, red hair, green eyes, black horns, eyeliner, mascara, red lips, curvy, narrow waist, ... an other outfit ...</code></li></ul>She'll be right here 😉This is the result of my first "large" local training at 1536 resolution using hires pictures without downscaling them (the dataset is going to be uploaded for the curious): 349 pictures at 2 Megapixels, 12 epochs, ~13h of training for a 128 dim (64 alpha) LoRA i merged in anima-base-1.0.<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/ec856696-2f92-41ec-abb2-837ba3bf31ba/original=true/ec856696-2f92-41ec-abb2-837ba3bf31ba.jpeg" />This will be a base for my tests of LoRA and stuff, i'll see if i "improve" it further later. Of course, it works very nicely with the <a target="_blank" rel="ugc" href="https://civarchive.com/models/2619830/turbo-for-anima-less-steps">Turbo LoRA</a>Thanks for reading! 💜<hr />Legal<span style="background-color:rgb(26, 27, 30);color:rgb(141, 141, 143);font-family:-apple-system, BlinkMacSystemFont, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:16px">The CircleStone Model is licensed by CircleStone Labs LLC under the CircleStone Non-Commercial License. Copyright CircleStone Labs LLC. IN NO EVENT SHALL CIRCLESTONE LABS LLC BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH USE OF THIS MODEL.

Lianima - Personal ANIMA base

Posted on Civitai, upload found elsewhere are FORBIDDEN and done without the author permission.

Files with SHA256: 0b1b886cb55c452fa8375b91af6af40c499642cff6660c7b65aefaf30615ff6e