<h3 id="datasets-detail:">Datasets Detail:</h3><ul><li><p>67 clips in 256x256 24FPS edited with Davinci Resolve</p></li></ul><h3 id="training-detail:">Training Detail:</h3><ul><li><p>Using L40S Runpod</p></li><li><p>Takes around 14-16 hours</p></li><li><p>Using Diffusion-pipe</p></li><li><p>32 epoch</p></li><li><p>Around 3k steps</p></li><li><p>Testing around 10 hours</p></li><li><p>Thx to this tutorial: <a target="_blank" rel="ugc" href="https://civarchive.com/articles/11942/training-a-lora-the-right-way">https://civarchive.com/articles/11942/training-a-lora-the-right-way</a></p></li><li><p>Use model I2V 480p</p></li><li><p>frame_buckets = [1,25,49,73,97,121]</p></li></ul><h3 id="good-at:">Good at:</h3><ul><li><p>Reverse and Stand Carry Position</p></li></ul><ul><li><p>or any pose that describe Up &amp; Down thrust/movements/motions</p></li><li><p>realism styles</p></li><li><p>fp16 text encoder (<a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files/text_encoders">https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files/text_encoders</a>)</p></li></ul><h3 id="bad-at:">Bad at:</h3><ul><li><p>Back and forth movements</p></li><li><p>Anime styles</p></li></ul><h3 id="how-to-use:">How to use:</h3><p>Since i trained with natural language you can describe it naturally as long includes the trigger word <strong>r3turnth15rsacp, </strong>tho when i tested it you can still using without it.</p><pre><code>r3turnth15rsacp, ...[Describe your image by your own imagination]...</code></pre><pre><code>OR JUST LOOK AT THE ANIMATIONS I'VE POSTED</code></pre><h3 id="workflow:">Workflow:</h3><p><a target="_blank" rel="ugc" href="https://mega.nz/file/et10xLZL#sR-0GtDPgb-F_98AZ9IVjw7RMBAW5_VIn-ZRfnnhO5A">https://mega.nz/file/et10xLZL#sR-0GtDPgb-F_98AZ9IVjw7RMBAW5_VIn-ZRfnnhO5A</a></p><p></p><h3 id="i-need-feedback-now!">I NEED FEEDBACK NOW!</h3><p></p>

Files with SHA256: c6687ec412ba440886f393fc21ac7bd7d28797b5c6d74e1ab594b2833df0c890