<h2 id="fibo2-finalcut">Fibo2 FinalCut</h2>Note: While Universal Format does not have comfy metadata it is still in comfy merged KQV format not diffusers<ul><li>Fully trained on diffusers (Split KQV)</li><li>Trained all parameters including time and noiserefiner</li><li>Trained on both text to image and edit task</li><li>Using a VAE like <a rel="ugc" href="https://civarchive.com/models/2661016/flux2-hdr-vae-krea2-flux2-moss">HDR VAE</a> will increase saturation on output.</li></ul><a target="_blank" rel="ugc" href="https://civarchive.com/articles/38307/the-fastest-edit-model-can-also-generate">Edit Workflow found here.</a><hr />Future Updates: I hope to include the massive WD14 training set I have into this at some-point however also training that task in edit takes 4x as much time.

Files with SHA256: 372c672124dc7f1d7ee520ca2db56d0a6fb0b0c4f24112fe4af1166732fc288f