[Character] Hanako/Multicostume (Blue Archive) LoRA/LoHa

<p><strong><span style="color:rgb(250, 82, 82)">Warning</span><span style="color:rgb(250, 176, 5)">⚠️</span><span style="color:rgb(250, 82, 82)">Please read below description due to nature of LoHa model behavior</span><span style="color:rgb(250, 176, 5)">⚠️</span></strong></p><p></p><h3 id="heading-25"><strong>If user apricate my work and wish to <span style="color:rgb(250, 82, 82)">Donate </span>or request for <span style="color:rgb(250, 82, 82)">Character Model Commission </span>,refer to my </strong><a target="_blank" rel="ugc" href="https://ko-fi.com/deatharms2010"><strong>Ko-fi</strong></a></h3><h3 id="heading-17">Thank you ^_^</h3><p></p><p><span style="color:rgb(193, 194, 197)">===========================================================</span></p><h2 id="heading-40"><br /><span style="color:#fab005">MODEL UPDATE:</span></h2><h3 id="heading-41"><span style="color:#fa5252">LoRA V1- (31/7/23)</span></h3><p>-Update verse5 office by increasing number of dataset. This should improve the accuracy of her design.<br /><br /></p><p><span style="color:rgb(193, 194, 197)">Reference from Blue Archieve Character, </span><a target="_blank" rel="ugc" href="https://bluearchive.fandom.com/wiki/Urawa_Hanako"><span style="color:rgb(193, 194, 197)">Hanako</span></a><span style="color:rgb(193, 194, 197)">. This model design to achieve multicostume with only short custom prompt.</span></p><p></p><p><strong>Simplify about this Model:</strong></p><p>-This Model quality has been <span style="color:rgb(250, 82, 82)">train enough to work on weight 1</span>.<br />-<em><span style="color:rgb(250, 176, 5)">Quality Prompt </span></em>such as "<span style="color:rgb(34, 139, 230)">Masterpiece </span>or <span style="color:rgb(34, 139, 230)">Best Quality</span>" is not necessary. This will <span style="color:rgb(250, 82, 82)">hammered the quality of image been generated</span>.</p><p>-User are expected to <span style="color:rgb(250, 82, 82)">use Highres with denoise 0.1~0.3</span> incase some checkpoint shown degrading quality of image.</p><p>-This Model been train with <strong><span style="color:rgb(250, 176, 5)">DAdaptAdam</span></strong>.</p><p>-The Model had been train enough to represent each of it Trigger Word except "<span style="color:rgb(34, 139, 230)">verse5 off1ce</span>" due to lack of sample image, this may rectified by next version.<br />-The <span style="color:rgb(250, 82, 82)">Halo may sometimes not appearing</span>, user can input normal prompt "<span style="color:rgb(34, 139, 230)">Halo</span>" instead.</p><p>-The model may look <span style="color:rgb(250, 82, 82)">desaturation due to Dadaptation optimizer</span>, user are advise to use <a target="_blank" rel="ugc" href="https://civitai.com/models/23906/kl-f8-anime2-vae"><strong>kl-f8-anime2</strong></a><strong> </strong>vae for clear saturation.</p><p></p><p><span style="color:rgb(193, 194, 197)">===========================================================</span><br /><br /><strong>How to use Lora:</strong><br /><span style="color:rgb(193, 194, 197)">LoCon &amp; LoHa (LyCoris):</span><br /><span style="color:rgb(193, 194, 197)">-If user had update to </span><span style="color:rgb(250, 82, 82)">latest SD1.5 or A1111, the Lora folder should supported Lycoris Model</span><span style="color:rgb(193, 194, 197)">, otherwise user may still follow below instruction: </span><br /><span style="color:rgb(193, 194, 197)">-Make sure your SD already contains LyCoris Extension </span><a target="_blank" rel="ugc" href="https://github.com/KohakuBlueleaf/a1111-sd-webui-lycoris-download">https://github.com/KohakuBlueleaf/a1111-sd-webui-lycoris<br /><span style="color:rgb(193, 194, 197)">-download</span></a><span style="color:rgb(193, 194, 197)"> the LoCon or LoHa models</span><br /><span style="color:rgb(193, 194, 197)">-Move the data into your Stable Diffusion -&gt; Models -&gt; Lycoris</span></p><p></p><p><span style="color:rgb(193, 194, 197)">===========================================================</span></p><p></p><p><strong>About Trigger Words:</strong></p><p>-This Model came with 3 type trigger words for 5 different costume, "<span style="color:rgb(34, 139, 230)">verse1 uniform</span>", "<span style="color:rgb(34, 139, 230)">verse2 swimsuit</span>", "<span style="color:rgb(34, 139, 230)">verse3 openshirt</span>", "<span style="color:rgb(34, 139, 230)">verse4 sport</span>", "<span style="color:rgb(34, 139, 230)">verse5 off1ce</span>".</p><p>-optionally use normal prompt "<span style="color:rgb(34, 139, 230)">Halo</span>" incase you want her original aesthetic. <br />-do not use "<span style="color:rgb(34, 139, 230)">Halo</span>" prompt if there already Halo on generate image otherwise you be seeing <span style="color:rgb(250, 82, 82)">2 Halo colliding with each other</span>.</p><p>-If user wish to <span style="color:rgb(250, 82, 82)">change to their desire outfit prompt</span>, <span style="color:rgb(250, 82, 82)">trigger word are not necessary</span>. <br /></p><p><span style="color:rgb(193, 194, 197)">===========================================================</span></p><p><br />Trigger Word Costume:<br /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/6a63cac7-a3ea-4068-b47d-f6032fbecfd5/width=525/6a63cac7-a3ea-4068-b47d-f6032fbecfd5.jpeg" /></p><p>Non-Trigger Word Costume:<br /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/429421c9-eefb-4700-9aba-ad5f6944a024/width=525/429421c9-eefb-4700-9aba-ad5f6944a024.jpeg" /></p>