Recursive Magic - Wan2.1 14B T2VThis t2v LoRA was trained on various fractal zooms, self-similarity and Droste effects and aims to help you replicate those dizzying experiences with custom items or persons.Because it's a t2v model, it works plug-and-play with LoRAs and VACE. The videos present are both with vanilla text2video and with VACE reference insertion. (4/1)All the videos here were created without loop forcing in the workflow - in form of flf2v or otherwise! (emergent ability?)This LoRA is quite experimental, and the dataset may be too diverse to focus fully. The anatomy can get really weird here, and it's maybe a feature. (Check out the fractal hands!) It also has a slight cool colors and 3D model look bias.There may be a version two in the future, though I think it's already looking dope enough.The trigger word is "fractal help".Some example prompts:<pre><code>fractal help. This animation shows an anime girl sitting still on a chair and holding a smartphone with an image of the anime girl herself sitting on a chair and holding a smartphone. The camera zooms in, showing the anime girl inside the smartphone holding another smartphone with the with an image of the anime girl herself sitting on a chair and holding a smartphone. The camera continues to zoom in, forming a perfect loop.</code></pre><pre><code>fractal help. This animation shows a wooden table standing on a gigantic wooden table. The table is holding a miniature version of itself. The camera zooms in showing this miniature table version holding even smaller miniature table version of itself. The camera zooms in further, completing a perfect loop. The background is a home interior.</code></pre>Leave a comment if you'd like to provide feedback or ask for usage tips.

Recursive Magic - Wan2.1 14B T2V

fractal-t2v.safetensors

Recursive Magic

v1.618

Recursive Magic - Wan2.1 14B T2VThis t2v LoRA was trained on various fractal zooms, self-similarity and Droste effects and aims to help you replicate those dizzying experiences with custom items or persons.Because it's a t2v model, it works plug-and-play with LoRAs and VACE. The videos present are both with vanilla text2video and with VACE reference insertion. (4/1)All the videos here were created without loop forcing in the workflow - in form of flf2v or otherwise! (emergent ability?)This LoRA is quite experimental, and the dataset may be too diverse to focus fully. The anatomy can get really weird here, and it's maybe a feature. (Check out the fractal hands!) It also has a slight cool colors and 3D model look bias.There may be a version two in the future, though I think it's already looking dope enough.The trigger word is "fractal help".Some example prompts:fractal help. This animation shows an anime girl sitting still on a chair and holding a smartphone with an image of the anime girl herself sitting on a chair and holding a smartphone. The camera zooms in, showing the anime girl inside the smartphone holding another smartphone with the with an image of the anime girl herself sitting on a chair and holding a smartphone. The camera continues to zoom in, forming a perfect loop.fractal help. This animation shows a wooden table standing on a gigantic wooden table. The table is holding a miniature version of itself. The camera zooms in showing this miniature table version holding even smaller miniature table version of itself. The camera zooms in further, completing a perfect loop. The background is a home interior.Leave a comment if you'd like to provide feedback or ask for usage tips.

Files with SHA256: 86161db587bd0323473449d076bb4332eaabb6c11d145e1e8b2531327b0c685f