ComfyUI-StyleTransferPlus

AesFAStyleBlend

Compact description unavailable for this node.

StyleTransferPlus · 3 inputs · 2 parameters · 1 output

AesPA

Compact description unavailable for this node.

StyleTransferPlus · 2 inputs · 2 parameters · 1 output

CAST

Download the CAST and UCAST models in the Test section (Google Drive) (Unzip). Download the vggnormalized.pth model in the Train section models/ │ .gitkeep │ vggnormalised.pth ─CASTmodel │ latestnetAE.pth │ latestnetDecA.pth # Safe to delete │ latestnetDecB.pth │ testopt.txt # Safe to delete │ └─UCASTmodel latestnetAE.pth latestnetDecA.pth # Safe to delete latestnetDecB.pth

StyleTransferPlus · 3 inputs · 0 parameters · 1 output

CoralColorTransfer

Compact description unavailable for this node.

StyleTransferPlus/Extra · 2 inputs · 0 parameters · 1 output

EFDM

Download the hmdecoderiter160000.pth in https://github.com/YBZh/EFDM/tree/main/ArbitraryStyleTransfer/models and put it in models/hmdecoderiter160000.pth Download vggnormalized.pth as instructed in CAST Arguments: - styleinterpweights: Weight for each style (only applicable for multiple styles). If empty, each style will have same weight (1/numstyles). Weights are automatically normalized so you can just enter anything - modelarch: ["adain", "adamean", "adastd", "efdm", "hm"], defaults to "efdm" - stylestrength: From 0.00 to 1.00. The higher the stronger the style. Defaults to 1.00 - docrop: Whether to resize and then crop the content and style images to size x size or just resize. Defaults to False - preservecolor: Whether to preserve the color of the content image. Defaults to False - size: Size of the height of the images after resizing. The aspect ratio of content is kept, while the styles will be resized to match content's height and width. The higher the size, the better the result, but also the more VRAM it will take. You may try usecpu to circumvent OOM. Tested with size=2048 on 12GB VRAM. - usecpu: Whether to use CPU or GPU for this. CPU takes very long! (default workflow = 21 seconds). Defaults to False.

StyleTransferPlus · 3 inputs · 6 parameters · 1 output

MicroAST

Download models from https://github.com/EndyWon/MicroAST/tree/main/models and place them in the models folder models/microast/contentencoderiter160000.pth.tar models/microast/decoderiter160000.pth.tar models/microast/modulatoriter160000.pth.tar models/microast/styleencoderiter160000.pth.tar Arguments: Similar to EFDM

StyleTransferPlus · 2 inputs · 2 parameters · 1 output

NeuralNeighbor

Arguments: - size: Either 512 or 1024, to be used for scaling the images. 1024 takes 6-12 GB of VRAM. Defaults to 512 - scalelong: Scale by the longest (or shortest side) to size. Defaults to True - flip: Augment style image with rotations. Slows down algorithm and increases memory requirement. Generally improves content preservation but hurts stylization slightly. Defaults to False - contentloss: Defaults to True > Use experimental content loss. The most common failure mode of our method is that colors will shift within an object creating highly visible artifacts, if that happens this flag can usually fix it, but currently it has some drawbacks which is why it isn't enabled by default [...]. One advantage of using this flag though is that [contentweight] can typically be set all the way to 0.0 and the content will remain recognizable. - colorize: Whether to apply color correction. Defaults to True - contentweight: weight = 1.0 corresponds to maximum content preservation, weight = 0.0 is maximum stylization (Defaults to 0.75) - maxiter: How long each optimization step should take. size=512 does 4 scaling steps maxiter. size=1024 does 5. The longer, the better and sharper the result. See more information on the original repository. No model is needed.

StyleTransferPlus · 2 inputs · 7 parameters · 1 output

TSSAT

Download the TSSAT-model.zip in the Model Testing section, and unzip it models/tssat/decoderiter160000.pth

StyleTransferPlus · 2 inputs · 3 parameters · 1 output

UniST

Download UniSTmodel.pt in Testing and vggr41.pth, decr41.pth in Training section and place them in models/unist/UniSTmodel.pt models/unist/decr41.pth models/unist/vggr41.pth The model only works with square images, so inputs are resized to size x size. docrop will resize height, then crop to square. The Video node is "more native" than the Image node for batched images (video) inputs. The model works with batch=3 (3 consecutive frames), so we split the video into such. | UniST | UniST Video | | :-----------------------------------------------------------: | :--------------------------------------------------------------: | | workflowunistimage.json | workflowunistvideo.json | | wfunistimage | wfmicroast | | | (Not a workflow-embedded image) |

StyleTransferPlus · 2 inputs · 2 parameters · 1 output

UniST_Video

Compact description unavailable for this node.

StyleTransferPlus · 2 inputs · 2 parameters · 1 output