<h2 id="gynecology-chair!"><span style="color:rgb(121, 80, 242)">Gynecology Chair</span>!?</h2><p></p><p>Yes indeed.</p><p></p><pre><code>A beautiful {Russian|French|Swedish|Swiss|African|Latina|Austrian|Dutch|English|American|Siberian|Chinese|Ukrainian|Thai|Czech|Japanese|Indian|Caucasian|Spanish|Portuguese} woman is sitting in a gynecology chair, ankles bound, wrists bound, legs in stirrups,

She has {very long|long|short} {blonde|brown|dirty-blonde|black|auburn} hair and is wearing a bikini,
{
A doctor is {putting a blindfold over her eyes|{sitting|standing} near her|securing her wrist to a support}||
}

indoor lighting, doctor's office, {high-angle|three-quarter|front} view</code></pre><p></p><h3 id="training-notes"><span style="color:rgb(250, 176, 5)">Training Notes</span></h3><p>WAN2.2 is a lot easier to train. I was really concerned about the time required to train two models and test two models but I am having lots of luck with shorter clips and the resulting models seem so much stabler than what I was used to with HunyuanVideo that epoch selection is taking me way less time.</p><p>I was exploring a single LORA file with a fellow user but it did come with slightly less quality, so two models for full quality it is!</p><p>My captioning style matches my wildcard prompts pretty closely in case you are curious.</p><p></p><p>A note about LTX2.3, as I've been getting requests for it. Everytime I read up on whether or not it's worth trying that out, I keep seeing that WAN2.2 is better, so... not planning on doing that anytime soon.</p><p></p><h3 id="disclaimer"><span style="color:rgb(253, 126, 20)">Disclaimer</span></h3><p>Ask your doctor!</p>

Files with SHA256: e99970f7e079610a9b585c2b2a1c6396acda5e22c3e187e7006e3fef42178233