<h2 id="intro-9svnvf2ou">Intro</h2>This is an HTML page that helps to see how a ckpt model KNOWS about specific TAGIf you don't see any clue about a certain TAG, that might indicate that this ckpt model doesn't handle the TAG well.<h2 id="tags-998i63mt5">Tags</h2>You can see tags from :<ul><li>(250412 updated) Face &amp; emotions<ul><li><a target="_blank" rel="ugc" href="https://danbooru.donmai.us/wiki_pages/tag_group%3Aface_tags">https://danbooru.donmai.us/wiki_pages/tag_group%3Aface_tags</a></li></ul></li><li>(250412 updated) Sex positions v2 ( with extra <code>sex</code> test tag ) <ul><li><a target="_blank" rel="ugc" href="https://danbooru.donmai.us/wiki_pages/tag_group%3Asexual_positions">https://danbooru.donmai.us/wiki_pages/tag_group%3Asexual_positions</a></li></ul></li><li>Camera related <ul><li><a target="_blank" rel="ugc" href="https://danbooru.donmai.us/wiki_pages/tag_group%3Aimage_composition">https://danbooru.donmai.us/wiki_pages/tag_group%3Aimage_composition</a></li></ul></li></ul><ul><li>Sex positions (v1)<ul><li><a target="_blank" rel="ugc" href="https://danbooru.donmai.us/wiki_pages/tag_group%3Asexual_positions">https://danbooru.donmai.us/wiki_pages/tag_group%3Asexual_positions</a></li></ul></li><li>Posture <ul><li><a target="_blank" rel="ugc" href="https://danbooru.donmai.us/wiki_pages/tag_group%3Aposture">https://danbooru.donmai.us/wiki_pages/tag_group%3Aposture</a></li></ul></li><li>Gestures <ul><li><a target="_blank" rel="ugc" href="https://danbooru.donmai.us/wiki_pages/tag_group%3Agestures">https://danbooru.donmai.us/wiki_pages/tag_group%3Agestures</a></li></ul></li></ul><h2 id="params-1ek6psu9g">Params</h2><ul><li>X<ul><li>seed : 1, 2, 3</li></ul></li><li>Y<ul><li>positive prompt : best quality, TAG</li><li>negative prompt : worst quality, bad quality, worst detail</li></ul></li></ul><h2 id="base-ckpt-model-used-c7sbvfpeg">Base ckpt model used</h2><ul><li>wai v13 -- WAI Illustrious v13</li></ul><h2 id="tips-6s3savpc5">Tips</h2><ul><li>Download &amp; unzip the zip file, find index.html and open<ul><li>You can click on image to preview</li><li>Size of cell could be toggle via the panel at the bottom right</li></ul></li><li>The script use for generating such benchmark grid would also be published later<ul><li>I don't like the XY tool in webui for benchmarking a LoRa that's trained. ( it requires a lot of manual work )<ul><li>So I wrote a tool to auto-run a benchmark based on pre-set config</li></ul></li><li>Any other tool you find useful, please recommend~! 🌷</li></ul></li></ul><h2 id="findings-0ej8bn1wy">Findings</h2><ul><li>1) I think some TAG would be triggered better with other related tag/s, instead of being 'TAG' alone by itself<ul><li>CLUE : spooning (sex position) should work, but none of the image give me a correct image (only 1 girl in the image). I think <code>1girl, 1boy, sex, spooning</code> will work better?</li></ul></li><li>2) The base model (wai i13) tends to draws 1girl by default when quality tags are given</li></ul>

Lab @7 : Sex Position, Posture, Gesture -- tags benchmark html page

lab7SexPositionPosture_WaiI13SexPositions.zip

[wai_i13] sex positions

[wai_i13] sex position v2

[wai_i13] emotions & face

[wai_i13] camera position

[wai_i13] camera angles

[wai_i13] gesture

[wai_i13] posture

[wai_i13] full collection

Files with SHA256: 62c0ddc931d1b0183539857d825afe7ebb04aba517452f48658c71f98529fc18