cowpuncher's picture
download
raw
3.18 kB
<h3>AI Reference Board</h3>
<p>Settings &rarr; Dockers &rarr; <b>AI Reference Board</b>. Works alongside the AI Image Diffusion plugin.</p>
<ol>
<li>Pick a folder, or type tags in the <b>Grab</b> box (like a prompt: <i>index finger raised, from side</i>) to pull
example pictures by tag. They're saved under <i>Pictures\AI Reference Board\&lt;tags&gt;</i>, each with a .txt of its tags.
A tag that finds nothing gets real tags close to it suggested, and a rare tag is pointed out: the models barely learned those.
<ul>
<li><b>From: Danbooru</b> needs no login. If Danbooru is down, Safebooru is used (safe pictures only).</li>
<li><b>From: Gelbooru</b> is the site Anima's tags follow. Its search only answers with an account's User ID and API key
(a free account will do): press <b>Gelbooru login ...</b>, open your Options page on gelbooru.com from there, and paste the
<i>API Access Credentials</i> line. Both values are kept as plain text in Krita's settings on your computer.</li>
<li>The rating box decides what is pulled: <b>General only</b> (the starting choice), <b>General + sensitive</b>,
<b>Any rating</b> or <b>Adult only</b>. The last two include adult pictures.</li>
</ul></li>
<li>Click a picture. Its tags show under the preview (<b>Copy tags</b> for the prompt). Drag a box on the preview to use
only that part; double-click (or <b>Whole image</b>) for all of it.</li>
<li>Choose how to use it and how strong, then <b>Add as reference</b>:
<ul>
<li><b>Reference / Style / Composition / Face</b>: image prompt for SDXL / Illustrious and edit styles.<br>
With <b>Anima</b> (needs the Anima Reference add-on): <b>Reference</b> = the character in the picture (two Reference pictures =
left and right character), <b>Style</b> = the character and the picture's art style, <b>Face</b> = one more picture of the same
character (up to 5 pictures in total). Composition is not available with Anima.</li>
<li><b>Pose (from picture)</b>: finds the pose in the picture and adds it as a Pose control layer, an editable stick figure.
Works on Anima too (needs the Anima pose control model).</li>
</ul></li>
</ol>
<p><b>Strength</b> is the same five-position slider the AI docker has on every control layer. A reference added from here keeps
that slider live in the AI docker, so you can move it there afterwards. Tick <b>Exact values</b> only if you want to type the
strength and range yourself: the AI docker then shows the control layer with <i>Use custom values</i> ticked and its slider
stays off until you untick that.</p>
<p>Pictures go on a hidden layer in an <i>[AI References]</i> group and are linked as control layers of AI Image Diffusion,
so they also show in the AI docker. Nothing lands on your canvas; the stick figure is left out of the picture itself.</p>
<p>Click a reference in <i>References in this document</i> to load its settings; change them and press <b>Apply</b>.
<b>On / Off</b> unlinks it but keeps it (On brings back the settings it had); <b>Remove</b> deletes it.</p>
<p>Whatever the source and rating: a picture rated above general is never kept when it is tagged as showing a child, and
such tags are only searched with <b>General only</b>.</p>

Xet Storage Details

Size:
3.18 kB
·
Xet hash:
ee97af9d5dd5de6bd66791ada4cfa51454b8b631bc4b229b38ef07c20ea97bae

Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.