Using this LORA, you can achieve a graceful spinning effect. Although it differs from my initial expectations, it is still functional and usable.Recommended usage:<ul><li>Keep the weight below 0.8 to effectively reduce unnatural body distortions.</li></ul>Over the past few days, I have conducted numerous experiments, testing various frame extraction methods and different frame combinations. However, I discovered that achieving better results requires more VRAM.Currently, I only have 24GB of VRAM, which limits my ability to create more complex LORAs. Through multiple combination tests, this version is the only one that remains within an acceptable range.Each combination requires several hours to process, and my GPU frequently loses CUDA response during video rendering. Perhaps I need more experimentation or collaboration with other creators to further improve this LORA.<hr />このLORAを使用すると、華麗なターン（回転）エフェクト を得ることができます。 最初に想定していたものとは少し違いますが、それでも十分に使えるレベルになっています。推奨設定:<ul><li>0.8以下の重み で使用すると、体の異常なねじれを効果的に抑えることができます。</li></ul>ここ数日間、さまざまな フレーム抽出方法 や 異なるフレームの組み合わせ を試しました。 しかし、より良い結果を得るためには、より大きなVRAMが必要 であることに気づきました。現在、私は 24GBのVRAM しか使用できないため、複雑なLORAを作成するのが難しい状況です。 さまざまな組み合わせを試しましたが、現時点で 正常な範囲内で動作するのはこのバージョンだけ です。各組み合わせの処理には 数時間 かかり、さらに 動画のレンダリング中にCUDAの応答が失われる ことも頻繁に発生しています。 より良い結果を得るために、さらに多くの試行錯誤を重ねるか、他のクリエイターと交流して知見を深める必要があるかもしれません。<hr />我本來打算做一個轉身後翹一下屁股然後勾一下腳的俏皮動作無奈target_frames 實在沒辦法吃到這個高的參數 可能要找找原作者還是大佬們有沒有解方

Graceful spinning effect  | Wan_video_14B ITV

Zandance_1.safetensors

0.5 | Wan_video_14B ITV

Created by allpleoleo439 on CivitaiFrom the Author----Using this LORA, you can achieve a graceful spinning effect. Although it differs from my initial expectations, it is still functional and usable.Recommended usage:<ul><li>Keep the weight below 0.8 to effectively reduce unnatural body distortions.</li></ul>Over the past few days, I have conducted numerous experiments, testing various frame extraction methods and different frame combinations. However, I discovered that achieving better results requires more VRAM.Currently, I only have 24GB of VRAM, which limits my ability to create more complex LORAs. Through multiple combination tests, this version is the only one that remains within an acceptable range.Each combination requires several hours to process, and my GPU frequently loses CUDA response during video rendering. Perhaps I need more experimentation or collaboration with other creators to further improve this LORA.

Files with SHA256: e1bebc5fc09c14a4bbee266c5f8f77830f414f3b735097a0b3a683b8c189f671