<h3 id="this-lora-aims-to-improve-the-fidelity-and-motion-of-pixellow-res-style-pov-paizuri-animations.-ze4hvmhkr">This lora aims to improve the fidelity and motion of pixel/low-res style POV paizuri animations.</h3>Works well with different hunyuan models, but if using a reduced VRAM model such as fp8 then I recommend an additional upscaler and/or refiner to achieve quality results. <a target="_blank" rel="ugc" href="https://civarchive.com/user/LatentDream">LatentDream</a> has an excellent <a target="_blank" rel="ugc" href="https://civarchive.com/models/1007385?modelVersionId=1268455">workflow</a> for this. Prompt and settings information is below.If you have questions, or want to commission a lora, join the <a target="_blank" rel="ugc" href="https://discord.gg/9Sd6aUEV9n">discord</a>.<h3 id="prompting:-pk11tom5b">Prompting:</h3>The trigger word is <code>povpixelpaizuri</code>, but here's a sample prompt to give you an idea of what to aim for. Keep in mind hunyuan works both with tags and phrases, so there's no reason to not take advantage and use both!<pre><code>povpixelpaizuri, 1girl, mature woman with huge breasts wearing fancy white dress giving paizuri, completely engulfing a penis between her breasts, male pov, pixel art, light-skinned female, light-skinned male, large breasts, clothed paizuri, paizuri under clothes, titfuck, tittyfuck, victorian mansion background, large penis, penis between breasts</code></pre><h3 id="recommended-settings:-or1yoeywh">Recommended Settings:</h3>Resolution: This lora was trained on 512px videos, and I have best results generating at that resolution. I've tried smaller 256px gens and gotten lucky, but your mileage may vary.Frames: The training videos were mostly between 2 and 3 seconds. I have best results generating 93 frames. Lower tends to create wacky wobbly motion, higher may look good but it takes so much longer to generate. Feel free to experiment and post your results. You can generate 1 frame to preview the art style, but it won't be at all representative of how the animation will look at higher frames. I'd only bother with a 1-frame preview if you want to double check that a character lora is working.Steps: 20-30. More steps will smooth out the animation and make it more believable.<h3 id="transparency:-swijb1bex">Transparency:</h3>This lora was trained on 13 videos of 512px resolution. It took $18 worth of Runpod credit on an L40s. I will never ask for your buzz or donations, but I will ask you to leave feedback in the comments so I can continue to improve my work.

Files with SHA256: ac9ae21945132ce7fe086da3bfdf4ce385765195ce337b56ba5e13d7bcc04159